Research

JAMA | Original Investigation

Long-Term Outcomes of Medical Management vs Bariatric Surgery

in Type 2 Diabetes
Anita P. Courcoulas, MD; Mary Elizabeth Patti, MD; Bo Hu, PhD; David E. Arterburn, MD; Donald C. Simonson, MD, ScD;
William F. Gourash, PhD; John M. Jakicic, PhD; Ashley H. Vernon, MD; Gerald J. Beck, PhD; Philip R. Schauer, MD;
Sangeeta R. Kashyap, MD; Ali Aminian, MD; David E. Cummings, MD; John P. Kirwan, PhD

Editorial page 643

IMPORTANCE Randomized clinical trials of bariatric surgery have been limited in size, type of Multimedia
surgical procedure, and follow-up duration.
Supplemental content
OBJECTIVE To determine long-term glycemic control and safety of bariatric surgery compared
with medical/lifestyle management of type 2 diabetes.

DESIGN, SETTING, AND PARTICIPANTS ARMMS-T2D (Alliance of Randomized Trials of Medicine

vs Metabolic Surgery in Type 2 Diabetes) is a pooled analysis from 4 US single-center
randomized trials conducted between May 2007 and August 2013, with observational
follow-up through July 2022.

INTERVENTION Participants were originally randomized to undergo either medical/lifestyle

management or 1 of the following 3 bariatric surgical procedures: Roux-en-Y gastric bypass,
sleeve gastrectomy, or adjustable gastric banding.

MAIN OUTCOME AND MEASURES The primary outcome was change in hemoglobin A1c (HbA1c)
from baseline to 7 years for all participants. Data are reported for up to 12 years.

RESULTS A total of 262 of 305 eligible participants (86%) enrolled in long-term follow-up for
this pooled analysis. The mean (SD) age of participants was 49.9 (8.3) years, mean (SD) body
mass index was 36.4 (3.5), 68.3% were women, 31% were Black, and 67.2% were White.
During follow-up, 25% of participants randomized to undergo medical/lifestyle management
underwent bariatric surgery. The median follow-up was 11 years. At 7 years, HbA1c decreased
by 0.2% (95% CI, −0.5% to 0.2%), from a baseline of 8.2%, in the medical/lifestyle group and
by 1.6% (95% CI, −1.8% to −1.3%), from a baseline of 8.7%, in the bariatric surgery group.
The between-group difference was −1.4% (95% CI, −1.8% to −1.0%; P < .001) at 7 years and
−1.1% (95% CI, −1.7% to −0.5%; P = .002) at 12 years. Fewer antidiabetes medications were
used in the bariatric surgery group. Diabetes remission was greater after bariatric surgery
(6.2% in the medical/lifestyle group vs 18.2% in the bariatric surgery group; P = .02) at 7
years and at 12 years (0.0% in the medical/lifestyle group vs 12.7% in the bariatric surgery
group; P < .001). There were 4 deaths (2.2%), 2 in each group, and no differences in major
cardiovascular adverse events. Anemia, fractures, and gastrointestinal adverse events were
more common after bariatric surgery.

CONCLUSION AND RELEVANCE After 7 to 12 years of follow-up, individuals originally randomized

to undergo bariatric surgery compared with medical/lifestyle intervention had superior glycemic
control with less diabetes medication use and higher rates of diabetes remission.

TRIAL REGISTRATION ClinicalTrials.gov Identifier: NCT02328599

Author Affiliations: Author

affiliations are listed at the end of this
article.
Corresponding Author: Anita P.
Courcoulas, MD, Department of
Surgery, University of Pittsburgh,
UPMC Magee-Womens Hospital, 300
Halket St, Ste C-400, Pittsburgh, PA
JAMA. 2024;331(8):654-664. doi:10.1001/jama.2024.0318 15213 ([email protected]).

654 (Reprinted) jama.com

© 2024 American Medical Association. All rights reserved.

Downloaded from jamanetwork.com by College of Medicine of Seoul National Univ. user on 03/23/2024
 Long-Term Outcomes of Medical Management vs Bariatric Surgery in Type 2 Diabetes Original Investigation Research

T
ype 2 diabetes is a progressive multifactorial disease af-
fecting an estimated 10% of the world’s population, or Key Points
more than 500 million adults.1 In the US, type 2 diabe-
Question What is the long-term durability of glycemic control and
tes accounts for $237 billion in direct medical costs and safety of bariatric surgery compared with medical/lifestyle
$90 billion in lost productivity annually, while care for people management of type 2 diabetes?
with diabetes accounts for 1 in 4 US health care dollars.2 Sev-
Findings Bariatric surgery led to superior glycemic control
eral small randomized clinical trials (RCTs) and larger obser-
compared with medical/lifestyle intervention (between-group
vational studies indicate that bariatric surgery is superior to difference in hemoglobin A1c of 1.4% at 7 years and 1.1% at 12
medical and lifestyle therapies for treatment of type 2 years), with less diabetes medication usage and higher rates of
diabetes.3-5 To date, RCTs have been limited in number, sample diabetes remission.
size, single site, operation type, severity of obesity, and
Meaning These results combined with existing evidence support
follow-up duration. Thus, despite the growing body of evi- the use of bariatric surgery for treatment of type 2 diabetes in
dence, most clinicians and payers do not recommend bariat- people with obesity.
ric surgery for type 2 diabetes unless an individual has a body
mass index (BMI) of 35 or higher,6 and less than 1% of those
with a BMI of 35 or higher consider or pursue surgical Program and Look AHEAD interventions. Patients who
treatment.7 Moreover, despite the recent advent of medica- underwent an operation had standard postoperative care,
tions that achieve a degree of weight loss approaching that of which included guidance on eating and activity, at least
some bariatric surgical procedures, these agents are costly, do monthly, in the first year. Observational follow-up was con-
not have proven long-term efficacy, and require long-term use ducted through July 2022. Data from extended follow-up
to maintain weight loss. were pooled and harmonized across sites for this analysis.
The Alliance of Randomized Trials of Medicine vs Additional details are provided in Supplement 2.
Metabolic Surgery in Type 2 Diabetes (ARMMS-T2D) consor-
tium pooled long-term observational results from 4 US Primary and Secondary Outcomes
single-center randomized trials to determine the efficacy, The primary outcome was the between-group difference
durability, and safety of bariatric surgery compared with in the percent change in HbA1c from baseline to 7 years. Addi-
medical/lifestyle treatment for type 2 diabetes. 8 To our tionally, data up to 12 years are reported for participants who
knowledge, this represents the largest pooled analysis with reached that follow-up point before study closure. Second-
the longest follow-up to date. We previously reported results ary outcomes included change over time in HbA1c, HbA1c less
from ARMMS-T2D at 3 years, demonstrating that bariatric than 7.0%, and diabetes remission (HbA1c <6.5% without dia-
surgery is more effective and durable than medical/lifestyle betes medications for at least 3 months, assessed annually).21,22
intervention for type 2 diabetes remission, even in those with Other secondary outcomes were between-group differences
a BMI of 25 to less than 35.9 Herein, we report long-term in change in weight, BMI, lipid levels, blood pressure, medi-
follow-up of participants in ARMMS-T2D at the primary end cation use, major adverse cardiovascular events, and micro-
point of 7 years, and up to 12 years, after randomization. vascular complications. Main results are presented by
randomized group as well as by surgical procedure type.
For all outcomes, we hypothesized better results for bariatric
surgery vs medical/lifestyle treatment. Throughout, to con-
Methods vert HbA 1 c to mmol/mol, use the following equation:
Study design 10.93 × HbA1c − 23.50.
Randomized trials were conducted from May 1, 2007, to August Adverse events related to the original interventions were
30, 2013, in the US at 4 centers: Cleveland Clinic (STAMPEDE), collected during annual study visits and at quarterly tele-
Joslin Diabetes Center/Brigham and Women’s Hospital phone calls by patient self-report, with verification by medi-
(SLIMM-T2D), University of Pittsburgh (TRIABETES), and cal record review. Serious adverse events were defined as
University of Washington/Kaiser Permanente Washington those resulting in death, a life-threatening event, hospitaliza-
(CROSSROADS).10-20 Written consent was obtained from par- tion, significant disability, incapacity, or need for urgent
ticipants at each center, ethics approval was obtained to pool medical or surgical intervention to prevent a serious out-
study results, and the pooled study analysis was registered at come. Events with specific relevance, including revisional
clinicaltrials.gov (NCT02328599). The protocol is provided metabolic/bariatric procedures, severe hyperglycemia or
in Supplement 1. Race and ethnicity were included in each hypoglycemia, and nutritional deficiencies are reported
study because they may contribute to differential outcomes up to 12 years.
from the interventions. Participants self-identified their race
and ethnicity based on fixed categories. Original study eligi- Statistical Analysis
bility included diagnosis of type 2 diabetes, BMI of 27 to 45, The primary statistical analysis followed the intention-to-
and age of 18 to 65 years. Randomization protocol and opera- treat principle comparing the change in HbA1c from baseline
tion types differed by site. Although medical/lifestyle inter- to 7 years between the 2 groups. Percent change in HbA1c was
ventions differed by site, they were all intensive compared assessed using a linear mixed-effect model, which included
with usual care and were based on the Diabetes Prevention group, visit, group × visit interaction, site, and baseline HbA1c

jama.com (Reprinted) JAMA February 27, 2024 Volume 331, Number 8 655

© 2024 American Medical Association. All rights reserved.

Downloaded from jamanetwork.com by College of Medicine of Seoul National Univ. user on 03/23/2024
 Research Original Investigation Long-Term Outcomes of Medical Management vs Bariatric Surgery in Type 2 Diabetes

cohort. All analyses were performed using RStudio (R Foun-

Figure 1. Assembly of the Trials in the Alliance of Randomized Trials
of Medicine vs Metabolic Surgery in Type 2 Diabetes (ARMMS-T2D)
dation) and SAS version 9.4 (SAS Institute).

355 Patients randomized in 4 original trials

155 In STAMPEDE
88 In SLIMM-T2D Results
69 In TRIABETES
43 In CROSSROADS The 4 original trials included 355 individuals with type 2 dia-
betes who were randomized to undergo bariatric surgery vs
39 Withdrew before intervention medical/lifestyle intervention for management of type 2 dia-
betes (Figure 1).8 Randomization occurred at time of entry into
316 Eligible for ARMMS-T2D trial the original trials; 39 individuals withdrew before the inter-
145 From STAMPEDE vention (more than half due to dissatisfaction with assigned
78 From SLIMM-T2D
61 From TRIABETES
treatment group), leaving 316 eligible for the pooled analysis.
32 From CROSSROADS Between the end of the original trials and the beginning of the
pooled analysis, 9 participants withdrew their consent and 2
14 Excluded died. Of the 305 individuals available, 262 (86%) enrolled
12 Withdrew or lost to (Figure 1). Baseline characteristics of participants in the
follow-upa
2 Died medical/lifestyle (n = 96) and surgical (n = 166) groups are
shown in Table 1. Mean (SD) age was 49.9 (8.3) years, 68.3%
of participants were women, 31% were Black, and 67.2% were
305 Available for long-term
follow-up and randomized White. Mean (SD) BMI was 36.4 (3.5) (eFigure 1 in Supple-
ment 2) and 96 participants (36.6%) had a baseline BMI less
than 35. Mean (SD) baseline HbA1c was 8.5% (1.5%). Median
193 Randomized to undergo 122 Randomized to undergo (range) follow-up was 11 (7-15) years; some participants with
bariatric surgery medical/lifestyle intervention
106 Roux-en-Y gastric bypass later enrollment did not reach 12 years of follow-up. A com-
49 Sleeve gastrectomy parison of patient characteristics with 7 and 12 years of
38 Adjustable gastric banding
follow-up is shown in eTable 1 in Supplement 2.
Absolute values for HbA1c over 12 years are shown in
166 Enrolled in ARMMS-T2D 96 Enrolled in ARMMS-T2D
Figure 2A. Despite higher baseline values, the bariatric sur-
gery group had significantly lower HbA1c levels than the medi-
a
Three participants who were lost to follow-up in the original trials were
cal/lifestyle group at all points after baseline (P < .001;
successfully rerecruited into ARMMS-T2D.8
Figure 2A). At 7 years, mean HbA1c decreased to 8.0% from a
baseline of 8.2% (difference, 0.2% [95% CI, −0.5% to 0.2%])
in the medical/lifestyle group and from 8.7% to 7.2% (differ-
as fixed effects, as well as a random intercept at the partici- ence, 1.6% [95% CI, −1.8% to −1.3%]) in the bariatric surgery
pant level. Least-square means were derived at each visit by group (Table 2). The between-group difference was −1.4%
group, and the group comparison was performed by applying (95% CI, −1.8% to −1.0%; P < .001) at 7 years (Table 2) and −1.1%
appropriate linear contrast to the model. A per-protocol sen- (95% CI, −1.7% to −0.5%; P = .002) at 12 years (eTable 2 in
sitivity analysis was also conducted to account for crossover Supplement 2). At year 7, there were no significant differ-
from the medical/lifestyle group to the surgical group using ences in the improvements in mean HbA1c from baseline be-
the inverse probability weighting approach,23-25 where the tween Roux-en-Y gastric bypass (difference, −1.7% [95% CI,
model for the inverse probability weight assumed the current −2.0% to −1.3%]) and sleeve gastrectomy (difference, −2.0%
treatment depended on the previous treatments and out- [95% CI, −2.6% to −1.5%]); HbA1c improvement after adjust-
comes. Missing clinical or laboratory data were recovered using able gastric banding (−0.8% [95% CI, −1.3% to −0.2%]) was less
the participants’ electronic medical record, as described in than for sleeve gastrectomy (P = .007) and Roux-en-Y gastric
Supplement 2. Death or loss to follow-up were assumed to be bypass (P = .03) (Figure 2B). A per-protocol sensitivity analy-
random censoring events, and the monotone missing data sis accounting for the 25% crossover from medical/lifestyle in-
caused by these events were also considered missing at ran- tervention to a surgical intervention showed a change in mean
dom. Further details about handling of missing data are pro- HbA 1c of 0.1% (95% CI, −0.5% to 0.7%) for the medical/
vided in Supplement 2. Binary outcomes, such as diabetes re- lifestyle group and −1.4% (95% CI, −1.7% to −1.2%) for the bar-
mission and use of medications for diabetes, were summarized iatric surgery group at 7 years, with a between-group differ-
as percentages and analyzed using the generalized estimat- ence of −1.5% (95% CI, −2.1% to −0.9%; P < .001; eFigure 2A
ing equation approach controlling for site and baseline HbA1c; in Supplement 2).
standard errors of parameter estimates were obtained from the Overall, 0.5% of participants in the medical/lifestyle
sandwich method. The results of secondary outcomes were not group achieved remission of diabetes at 1 year, compared
adjusted for multiple testing. Exploratory analyses were per- with 50.8% in the bariatric surgery group (Figure 3). Over
formed to investigate the subgroup of participants with BMI time, the percentage achieving remission decreased in the
less than 35 using the same set of models as for the overall bariatric surgery group, but remission rates remained higher

656 JAMA February 27, 2024 Volume 331, Number 8 (Reprinted) jama.com

© 2024 American Medical Association. All rights reserved.

Downloaded from jamanetwork.com by College of Medicine of Seoul National Univ. user on 03/23/2024
 Long-Term Outcomes of Medical Management vs Bariatric Surgery in Type 2 Diabetes Original Investigation Research

Table 1. Baseline Characteristics of Participants by Group (N = 262)

Bariatric surgery type

Roux-en-Y Sleeve Adjustable
Medical/lifestyle Bariatric surgery gastric bypass gastrectomy gastric banding
Characteristic (n = 96) (n = 166) (n = 89) (n = 41) (n = 36)
Demographics, No. (%)
Age, y 51.4 (6.8) 49.0 (9.0) 49.1 (9.0) 48.3 (7.7) 49.6 (10.3)
Sex
Women 62 (64.6) 117 (70.5) 61 (68.5) 32 (78.0) 24 (66.7)
Men 34 (35.4) 49 (29.5) 28 (31.5) 9 (22.0) 12 (33.3)
Race
Black 35 (36.5) 46 (27.7) 23 (25.8) 13 (31.7) 10 (27.8)
White 59 (61.5) 118 (71.1) 64 (71.9) 28 (68.3) 26 (72.2)
Othera 2 (2.1) 2 (1.2) 2 (2.3) 0 0
Anthropometrics, mean (SD)
Waist, cm 113.7 (9.6) 115.0 (9.9) 116.1 (9.9) 113.3 (10.2) 114.5 (9.7)
[n = 95]
Weight, kg 105.6 (15.5) 103.5 (15.3) 105.2 (15.3) 100.2 (16.7) 103.1 (13.0)
BMI 36.2 (3.4) 36.6 (3.6) 37.0 (3.4) 36.3 (4.2) 35.9 (3.2)
BMI <35 40 (41.7) 56 (33.7) 26 (29.2) 15 (36.6) 15 (41.7)
Systolic BP, mm Hg 129.7 (15.8) 134.4 (17.7) 135.0 (18.4) 135.8 (19.9) 131.6 (12.7)
Diastolic BP, mm Hg 79.5 (9.6) 80.4 (10.0) 80.7 (9.8) 81.9 (12.2) 78.2 (7.2)
Diabetes duration, y 8.8 (5.2) 8.3 (5.5) 8.8 (5.9) 7.8 (4.6) 7.5 (5.1)
Laboratory
HbA1c, mean (SD), % 8.2 (1.2) 8.7 (1.7) 8.7 (1.6) 9.4 (1.6) 8.2 (1.8)
HbA1c <7.0%, No. (%) 11 (11.5) 20 (12.0) 9 (10.1) 0 11 (30.6)
Fasting glucose, 156.5 (50.0) 172.0 (69.7) 171.0 (69.5) 172.1 (66.1) 174.6 (75.9)
mean (SD), mg/dL [n = 95]
Total cholesterol, 172.6 (38.5) 179.6 (44.8) 176.1 (40.7) 191.2 (46.8) 174.9 (50.7)
mean (SD), mg/dL [n = 95]
HDL, mean (SD), 44.3 (13.2) 42.9 (11.6) 43.6 (11.8) 45.5 (12.2) 38.1 (9.0)
mg/dL [n = 95] Abbreviations: ACEi,
LDL, mean (SD), 96.3 (33.2) 100.3 (34.2) 96.3 (31.8) 110.8 (41.0) 97.8 (28.7) angiotensin-converting enzyme
mg/dL [n = 94] [n = 159] [n = 88] [n = 30] inhibitor; ARB, angiotensin receptor
Triglycerides, mean 140.0 145.0 143.0 160.0 142.5 blocker; BMI, body mass index;
(SD), mg/dL (92.5-221.5) (103.0-231.8) (100.0-239.0) (120.0-214.0) (94.0-251.5) BP, blood pressure; HDL,
[n = 95] high-density lipoprotein cholesterol;
Serum creatinine, 0.7 (0.2) 0.7 (0.2) 0.7 (0.2) 0.7 (0.2) 0.8 (0.2) LDL, low-density lipoprotein
mean (SD), mg/dL [n = 95] cholesterol.
Urine 6.0 (4.0-14.5) 9.0 (4.0-22.5) 7.5 (3.0-28.0) 9.0 (7.0-22.0) 6.0 (3.8-12.2) SI conversion factors: To convert
albumin:creatinine [n = 74] [n = 119] [n = 62] [n = 16] hemoglobin A1c (HbA1c) to mmol/mol,
ratio, mean (SD)
multiply by 10.93 and subtract 23.50;
Medication use, No. (%) glucose to mmol/L, multiply by
Statins 71 (74.0) 121 (72.9) 66 (74.2) 31 (75.6) 24 (66.7) 0.0555; cholesterol to mmol/L,
multiply by 0.0259; triglycerides to
ACEi/ARB 65 (67.7) 108 (65.1) 63 (70.8) 25 (61.0) 20 (55.6)
mmol/L, multiply by 0.0113;
ACEi 59 (61.5) 78 (47.0) 47 (52.8) 17 (41.5) 14 (38.9) creatinine to μmol/L, multiply by
Insulin 36 (37.5) 82 (49.4) 46 (51.7) 17 (41.5) 19 (52.8) 88.4.
a
β-Blockers 16 (16.7) 30 (18.2) 15 (16.9) 6 (14.6) 9/35 (25.7) Including 2 patients identifying as
Asian race and 2 patients who
ARB 8 (8.3) 30 (18.1) 16 (18.0) 8 (19.5) 6 (16.7)
reported more than 1 race.

than in the medical/lifestyle group. At year 7, remission was of participants in the bariatric surgery group (odds ratio, 3.2
6.2% in the medical/lifestyle group compared with 18.2% in [95% CI, 1.8-5.9]; P < .001; Table 2). Similar differences were
the bariatric surgery group (odds ratio, 3.4 [95% CI, 1.3-9.2]; observed using the threshold of HbA1c less than or equal to
P = .02; Figure 3), with the difference remaining statistically 6.5% (P = .002; Table 2). Moreover, improved glycemic con-
significant at 12 years (P < .001). Rates of remission were trol in the bariatric surgery group was achieved using fewer
24.5% in the Roux-en-Y gastric bypass subgroup, 15.2% in the medications. Rates of diabetes medication use at baseline
sleeve gastrectomy group, and 8.9% in the adjustable gastric were similar between the groups (Table 2; eFigure 3 in
banding group; comparisons between surgical procedures Supplement 2) and did not change significantly over time
were not statistically significant. in the medical/lifestyle group (P = .19 at year 7 and 0.12 at
At 7 years, 26.7% of participants in the medical/lifestyle year 12); in contrast, medication usage was reduced from
group had an HbA1c less than 7.0%, compared with 54.1% 97.6% (162/166) at baseline to 38.0% (62/163) at year 1 in the

jama.com (Reprinted) JAMA February 27, 2024 Volume 331, Number 8 657

© 2024 American Medical Association. All rights reserved.

Downloaded from jamanetwork.com by College of Medicine of Seoul National Univ. user on 03/23/2024
 658
Figure 2. HbA1c and Weight Loss by Group and Procedure Type

A Hemoglobin A1c (HbA1c) by group B HbA1c by bariatric surgical procedure

14 Bariatric surgery Medical/lifestyle 14 Adjustable gastric banding Roux-en-Y gastric bypass Sleeve gastrectomy

12 12

10 10
Research Original Investigation

HbA1c, %
HbA1c, %
8 8

6 6

4 4
0 1 2 3 4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 12
Annual visit Annual visit

JAMA February 27, 2024 Volume 331, Number 8 (Reprinted)

No. at risk No. at risk
Bariatric surgery 166 164 160 157 147 152 118 136 119 126 119 100 83 Adjustable gastric banding 36 35 33 34 34 30 29 33 22 24 23 10 16
Medical/lifestyle 96 92 88 86 80 86 78 82 72 71 68 55 31 Roux-en-Y gastric bypass 89 88 86 84 75 82 60 70 65 68 63 58 40
Sleeve gastrectomy 41 41 41 39 38 40 29 33 32 34 33 32 27

C Weight loss D Weight loss by bariatric surgical procedure

20 20

10 10

0 0

Downloaded from jamanetwork.com by College of Medicine of Seoul National Univ. user on 03/23/2024
–10 –10

–20 –20

Weight loss, %
Weight loss, %

–30 –30

© 2024 American Medical Association. All rights reserved.

–40 –40

–50 –50
0 1 2 3 4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 12
Annual visit Annual visit
No. at risk No. at risk
Bariatric surgery 166 164 161 158 144 149 122 139 121 126 121 106 85 Adjustable gastric banding 36 34 33 32 31 29 32 31 22 24 25 12 18
Medical/lifestyle 96 91 84 86 79 78 77 75 73 73 70 60 34 Roux-en-Y gastric bypass 89 89 87 86 75 81 60 74 64 68 62 61 41
Sleeve gastrectomy 41 41 41 40 38 39 30 34 35 34 34 33 26

The lines and dots represent the least-square estimates obtained from the model and the boxplots represent the tops and bottoms of the boxes represent the IQR, and the whiskers represent the highest and lowest values
the raw data. Horizonal lines within the boxes demonstrate median values, dots indicate mean values, within 1.5 × the IQR.

jama.com
Long-Term Outcomes of Medical Management vs Bariatric Surgery in Type 2 Diabetes
 Long-Term Outcomes of Medical Management vs Bariatric Surgery in Type 2 Diabetes Original Investigation Research

Table 2. Laboratory and Clinical Outcomes at Year 7 and Difference From Baseline a

Medical/lifestyle Bariatric surgery Group comparison

Baseline Year 7 Change Baseline Year 7 Change Difference
Outcome (n = 96) (n = 82) (95% CI)b (n = 166) (n = 136) (95% CI) in changec P value
Primary outcome
HbA1c, mean (SD), % 8.2 (1.2) 8.0 (1.8) −0.2 (−0.5 to 0.2) 8.7 (1.7) 7.2 (1.4) −1.6 (−1.8 to −1.3) −1.4 (−1.8 to −1.0) <.001
Secondary outcomes
Fasting glucose, 156.5 144.6 −3.8 172.0 125.1 −14.1% −10.3% .13
mean (SD), mg/dLd (50.0) (57.3) (−14.8% to 7.2%) (69.7) (47.0) (−22.0% to −6.3%) (−23.6% to 2.9%)
Weight, 105.6 96.2 −8.3% 103.5 83.6 −19.9% −11.6% <.001
mean (SD), kg (15.5) (16.6) (−10.5% to −6.1%) (15.3) (15.8) (−21.6% to −18.1%) (−14.3% to −8.9%)
SBP, mean (SD), 129.7 128.7 −1.1% 134.4 128.6 −3.4% −2.3% .19
mm Hg (15.8) (15.7) (−3.9% to 1.7%) (17.7) (15.3) (−5.6% to −1.2%) (−5.8% to 1.1%)
DBP, mean (SD), 79.5 (9.6) 74.6 (9.7) −4.3% 80.4 (10) 74.3 (10.4) −6.0% −1.7% .32
mm Hg (−7% to −1.6%) (−8.1% to −3.8%) (−5.0% to 1.7%)
LDL, mean (SD), 96.3 (33.2) 97.6 (36.6) 5.5% 100.3 103.1 10.8% 5.4% .34
mg/dL (−3.3% to 14.3%) (34.2) (36.4) (3.8% to 17.9%) (−5.6% to 16.3%)
HDL, mean (SD), 44.3 (13.2) 52.0 (17.0) 20.5% 42.9 (11.6) 56.5 (16.5) 37.4% 16.9% <.001
mg/dL (14.5% to 26.6%) (32.6% to 42.3%) (9.4% to 24.4%)
Total cholesterol, 172.6 171.7 −0.7% 179.6 (44.8) 181.4 4.9% 5.6% .07
mean (SD), mg/dL (38.5) (41.6) (−5.6% to 4.1%) (40.6) (1.0% to 8.7%) (−0.4% to 11.6%)
Triglycerides, 140 125 2.3% 144 107 −19.0% −21.3% .002
median (IQR), (92.5-221.5) (88-178.3) (−8.6% to 13.2%) (103 to 231) (82 to 142) (−27.8% to −10.2%) (−34.9% to −7.8%)
mg/dL
Serum creatinine, 0.7 (0.2) 0.8 (0.4) 9.5% 0.7 (0.2) 0.8 (0.2) 10.5% 1.1% .83
mean (SD), mg/dL (1.8% to 17.1%) (4.4% to 16.7%) (−8.4% to 10.5%)
Urine 6 (4-12) 8 (4-13.5) 1.3 (0.9 to 1.9) 9 (4.5 to 23) 6 (4 to 10) 0.9 (0.7 to 1.2) −0.4 (−1.0 to 0.1) .10
albumin:creatinine
ratio, median (IQR)
Remission of 0.6 6.2 10.4 (0.4 to 279.4) 0.6 18.2 36.2 (1.9 to 699.0) 3.39 (1.25 to 9.17) .02
diabetes, %
Use of medications 99.0 96.0 0.10 (0.01 to 3.27) 97.6 60.5 0.03 (0.01 to 0.11) 0.09 (0.03 to 0.24) <.001
for diabetes, %
Oral/GLP1 only 57.3 40.0 0.53 (0.29 to 0.97) 47.0 44.5 0.74 (0.46 to 1.20) 0.98 (0.53 to 1.82) .95
Insulin and/or 41.7 56.0 1.93 (1.07 to 3.46) 50.6 16.0 0.18 (0.11 to 0.31) 0.13 (0.06 to 0.29) <.001
oral/GLP1
HbA1c <7.0%, % 11.7 26.7 2.77 (1.38 to 5.54) 15.5 54.1 6.42 (3.63 to 11.4) 3.22 (1.76 to 5.88) <.001
HbA1c <6.5%, % 8.3 17.3 2.30 (1.19 to 4.47) 12.0 37.7 4.44 (2.46 to 8.01) 2.89 (1.48 to 5.64) .002
b
Abbreviations: DBP, diastolic blood pressure, GLP1, glucagon-like peptide 1; Net change presented for HbA1c, fold change (ie, ratio of year 7 data over
HDL, high-density lipoprotein cholesterol; LDL, low-density lipoprotein baseline data) for urine albumin:creatinine ratio, relative changes
cholesterol; SBP, systolic blood pressure. (ie, percentage change from baseline) for other numeric outcomes, and odds
SI conversion factors: To convert hemoglobin A1c (HbA1c) to mmol/mol, multiply ratios (7-year over baseline) for binary outcomes (remission, glycemic control,
by 10.93 and subtract 23.50; glucose to mmol/L, multiply by 0.0555; and use of medications for diabetes).
c
cholesterol to mmol/L, multiply by 0.0259; triglycerides to mmol/L, multiply by For numeric outcomes, the difference is defined as the 7-year change in the
0.0113; creatinine to μmol/L, multiply by 88.4. surgical group minus the 7-year change of the medical/lifestyle group; for
a
Mean (SD) or median (IQR) presented for baseline and year-7 data and binary outcomes, the difference is the odds ratio at year 7 (the odds of 7-year
least-square estimate (95% CI) presented for the changes and group outcome in the surgery group over the medical/lifestyle group).
d
comparisons. Only included fasting glucose measurements obtained at in-person study visits.

bariatric surgery group and remained significantly lower dur- Roux-en-Y gastric bypass than adjustable gastric banding
ing follow-up compared with baseline (60.5% [72/119] at year (22.7% vs 14.0%; P < .001; Figure 2D), but did not differ sig-
7; P < .001; Table 2; eFigure 3 in Supplement 2). Insulin usage nificantly in sleeve gastrectomy (19.7%) vs other surgical pro-
after bariatric surgery was significantly lower than in the cedures. At 12 years, the bariatric surgery group continued to
medical/lifestyle group (16% vs 56% at 7 years; P < .001; have superior weight loss (10.8% [95% CI, 8.2%-13.5%] in the
Table 2; eFigure 3 in Supplement 2) and the use of incretin/ medical/lifestyle group vs 19.3% [95% CI, 17.3%-21.3%] in the
glucagon-like peptide 1 agonist medications was higher in the bariatric surgery group; P < .001; Figure 2C). At 7 years, a BMI
medical/lifestyle group across all annual visits (P < .001; less than or equal to 25, indicating nonobesity, was achieved
eFigure 4 in Supplement 2). in 2.7% of participants in the medical/lifestyle group and
Weight loss trajectories up to 12 years are shown in 14.4% in the bariatric surgery group. At 12 years, these rates
Figure 2C and D. Weight loss was significantly greater 7 years were 0% in the medical/lifestyle group and 15.3% in the bar-
after bariatric surgery, with 8.3% weight loss (95% CI, 6.1%- iatric surgery group. eFigure 2B in Supplement 2 shows a
10.5%) in the medical/lifestyle group and 19.9% weight loss comparison of the intention-to-treat and per-protocol analy-
(95% CI, 18.1%-21.6%) in the surgical group (P < .001; ses of weight loss by group over time. At year 7, weight loss
Figure 2C). Weight loss at 7 years was significantly greater for was 5.6% for the medical/lifestyle group and 20.4% for the

jama.com (Reprinted) JAMA February 27, 2024 Volume 331, Number 8 659

© 2024 American Medical Association. All rights reserved.

Downloaded from jamanetwork.com by College of Medicine of Seoul National Univ. user on 03/23/2024
 Research Original Investigation Long-Term Outcomes of Medical Management vs Bariatric Surgery in Type 2 Diabetes

was 10.1% for the medical/lifestyle group and 19.3% for the bar-
Figure 3. Diabetes Remission
iatric surgery group, and the difference was 9.2% (95% CI, 5.6%-
60 12.9%), which was significantly different from the lower BMI
Participants achieving remission, % group (P = .03).
50 Medical/lifestyle Bariatric surgery
Deaths, major cardiovascular adverse events, and other
40 major adverse events are shown in Table 3. There were
4 deaths over 12 years: 2 in the medical/lifestyle group
30
(gunshot injury, disability from strokes leading to death)
20 and 2 in the bariatric surgery group (cardiac event, COVID-
19). Cardiovascular and other adverse events were similar
10
between the groups except for fractures, anemia (hemoglo-
0 bin <11.5 g/dL), and low iron (<59 μg/dL), which were more
0 1 2 3 4 5 6 7 8 9 10 11 12 common in the bariatric surgery group (Table 3; eTable 3 in
Annual visit Supplement 2). The bariatric surgery group had significantly
No. of participants
Medical/lifestyle 96 92 87 82 78 84 76 79 72 70 67 55 31
lower hemoglobin and higher vitamin B12 and vitamin D lev-
Bariatric surgery 166 164 151 149 140 146 108 131 116 125 117 99 82 els than the medical/lifestyle group (eTable 4 in Supple-
ment 2). Two of 96 participants (2.1%) in the medical/
Remission was defined as hemoglobin A1c less than 6.5% and not receiving any
lifestyle group initiated kidney dialysis during follow-up
medications for diabetes.
compared with none in the bariatric surgery group; 5 of 96
participants (5.2%) in the medical/lifestyle group compared
surgical group from the per-protocol analyses. At year 12, the with 2 of 166 (1.2%) in the bariatric surgery group experi-
per-protocol weight loss was 7.7% for the medical/lifestyle enced retinopathy (Table 3).
group and 19.4% in the bariatric surgery group after account- There were more gastrointestinal events, such as abdomi-
ing for crossovers. nal pain, dumping syndrome, and nausea/vomiting, in the bar-
There were no significant differences in systolic blood iatric surgery group than the medication/lifestyle group
pressure or low-density lipoprotein (LDL) cholesterol (eTable 5 in Supplement 2). Procedures (surgical, endo-
between the groups at 7 years (Table 2; eFigure 5A and 5C in scopic) during long-term follow-up are shown in eTable 6 in
Supplement 2). However, high-density lipoprotein (HDL) Supplement 2; there were no significant differences in proce-
cholesterol was significantly higher (P < .001; eFigure 5B in dures between the groups, except the combination of cross-
Supplement 2) and triglycerides were significantly lower overs, conversions, and revisions, which were more common
(P < .001; eFigure 5D in Supplement 2) over time in the bar- in the medical/lifestyle group. Specifically, 24 of 96 partici-
iatric surgery group. At year 7, the increase in HDL from base- pants (25%) in the medical/lifestyle group underwent bariat-
line was greater for bariatric surgery (20.5% vs 37.4%; differ- ric surgery (crossover) during follow-up (8 underwent Roux-
ence, 16.9% [95% CI, 9.4%-24.4%]; P < .001; Table 2). The en-Y gastric bypass, 15 underwent sleeve gastrectomy, and 1
percent change in triglycerides was 2.3% for medical/lifestyle underwent adjustable gastric banding) at a median (range) time
and −19.0% for bariatric surgery at 7 years (P = .002). There of 4.5 (0.4-9.8) years (eTable 7 in Supplement 2). In the bar-
were no significant changes in either serum creatinine or iatric surgery group, 15 participants (9%) underwent conver-
urine albumin-to-creatinine ratio between the groups sion or revisional operations, with 7 adjustable gastric band-
(Table 2). ing removals without further procedure, 4 adjustable gastric
Results of an exploratory analysis of the 96 participants banding conversions to Roux-en-Y gastric bypass, 1 adjust-
with a BMI of 27 to less than 35 vs those with a BMI of 35 or able gastric banding conversion to sleeve gastrectomy, and 3
greater at randomization is shown in eFigure 6 in Supple- sleeve gastrectomy revisions to Roux-en-Y gastric bypass (2 for
ment 2. The bariatric surgery group had significantly lower acid reflux, 1 for chronic fistula).
HbA1c levels than the medical/lifestyle group at all points
(P < .001; eFigure 6A in Supplement 2). At 7 years, the reduc-
tion in HbA1c in the lower BMI surgical subgroup was signifi-
cantly greater than in the lower BMI medical/lifestyle group
Discussion
(−0.4% vs −1.5%; difference, −1.2% [95% CI, −1.8% to −0.5%]; This study reports results from the largest and longest
eFigure 6A in Supplement 2). The reduction in HbA1c in the follow-up to date of individuals with type 2 diabetes and
higher BMI group was −0.1% for the medical/lifestyle group overweight/obesity randomized to receive medical/lifestyle
and −1.6% for the surgical group (difference, −1.5% [95% CI, vs surgical treatments. In this pooled analysis of patients
−2.1% to −1.0%]), which was not different from the lower BMI originally enrolled in 4 single center RCTs, at 7 years and up
group (P = .40). Likewise, percent weight loss at year 7 in the to 12 years of follow-up from randomization, bariatric surgery
subgroup of participants with a BMI less than 35 was signifi- was more effective and resulted in long-term improvement
cantly greater in the bariatric surgery group than the medical/ for glycemic control while using fewer medications and
lifestyle intervention group (5.6% vs 20.4%; difference, 14.8% weight loss. In parallel, increased rates of diabetes remission
[95% CI, 10.8%-18.8%]; P < .001; eFigure 6B in Supple- and reductions in medication use (including insulin) were
ment 2). Percent weight loss at year 7 in the higher BMI group observed in surgically treated participants. Bariatric surgery

660 JAMA February 27, 2024 Volume 331, Number 8 (Reprinted) jama.com

© 2024 American Medical Association. All rights reserved.

Downloaded from jamanetwork.com by College of Medicine of Seoul National Univ. user on 03/23/2024
 Long-Term Outcomes of Medical Management vs Bariatric Surgery in Type 2 Diabetes Original Investigation Research

was also more effective in improving HDL cholesterol and tri-

Table 3. Major Adverse Events and Events of Interest Through 12 Years
glycerides at 7 years and up to 12 years of follow-up. These
advantages should be balanced against the increased risk of No. (%)
nutritional deficiencies (anemia), gastrointestinal adverse Medical/lifestyle Bariatric surgery
Event (n = 96) (n = 166)
events, and bone fractures in long-term follow-up after bar- Deatha 2 (2.1) 2 (1.2)
iatric surgery.
Cardiovascular
Recent studies have highlighted that remission of diabe-
Coronary revascularization 7 (7.3) 15 (9)
tes is an important and achievable goal, particularly in early
Myocardial infarction 4 (4.2) 10 (6)
stages.26 These data demonstrate that bariatric surgery is su-
Unstable angina 2 (2.1) 4 (2.4)
perior to medical/lifestyle interventions, with remission
Significant arrhythmia 4 (4.2) 7 (4.2)
achieved in 51% of surgically-treated patients at 1 year and 18%
Heart failure 1 (1) 5 (3)
at 7 years, compared with 0.5% of patients in the medical/
Stroke/transient 3 (3.1) 5 (3)
lifestyle group at 1 year and 6% at 7 years.9 Participants in- ischemic attack
cluded in the analyses had long-standing and suboptimally con- Peripheral arterial disease 0 2 (1.2)
trolled type 2 diabetes (mean duration of diabetes, 8 years; Venous thromboembolism 2 (2.1) 1 (0.6)
mean HbA1c at randomization, 8.5%). By contrast, the medical/ Metabolic
lifestyle interventions, modeled after the Diabetes Preven-
Severe hypoglycemia 7 (7.3) 11 (6.6)
tion Program27 and Look AHEAD trials,28 were not sufficient
Diabetic ketoacidosis 1 (1) 0
to promote long-term improvements in glycemic control
Gastrointestinal
or achieve diabetes remission, despite modest weight loss. The
Gastric/anastomotic ulcer 2 (2.1) 10 (6)
relatively low remission rate at 7 years in the medical/
Bowel obstruction 1 (1) 3 (1.8)
lifestyle group is in contrast to the 2-year results from the
Gastrointestinal leaks 0 1 (0.6)
DiRECT trial.29 However, by design, participants in that study
Gallstones/cholecystitis 3 (3.1) 9 (5.4)
had a shorter duration of type 2 diabetes (mean, 3.1 years) and
Pancreatitis 1 (1) 3 (1.8)
were not treated with insulin, and follow-up in that study was
shorter. The reduction in remission over time after bariatric Alcohol-associated 0 2 (1.2)
cirrhosisb
surgical procedures has also been observed in other studies, Kidney
potentially related to weight regain, resolution of negative calo- 2 (2.1) 11 (6.6)
Kidney stones
rie balance, and progressive loss of β-cell function over time.
Initiation of dialysis 2 (2.1) 0
Hence, even among patients who experience postoperative dia-
Ocular
betes remission, continued surveillance for relapse is war-
Retinopathy 5 (5.2) 2 (1.2)
ranted. Even if relapse occurred, participants in the bariatric
Blindness 1 (1) 0
surgery group continued to experience better glycemic con-
Blood transfusion
trol while using fewer medications. Moreover, even rela-
For anemia 3 (3.1) 20 (12)
tively short-term remission has shown benefits for diabetes-
For gastrointestinal 2 (2.1) 5 (3)
related complications in other studies; risk of microvascular bleeding
disease has been estimated to be reduced by 19% for each year Miscellaneous
of achieved remission.30 Fracture 5 (5.2) 22 (13.3)
Among RCTs of surgical vs medical/lifestyle manage-
Cancerc 4 (4.2) 9 (5.4)
ment for type 2 diabetes, this analysis reports the longest
Suicide attempt 0 1 (0.6)
period of active observation, with follow-up through 12 years
a
The 2 deaths in the medical group were caused by gunshot injury and disability
after randomization. Beyond previous individual reports
following strokes and the 2 deaths in the surgical group were caused by a cardiac
from the ARMMS-T2D consortium, the longest-term RCTs event and COVID-19.
that reported comparison of surgical to medical/lifestyle b
One patient had metabolic dysfunction–associated steatohepatitis (MASH)
management of type 2 diabetes were studies by Ikramuddin with bridging fibrosis at the time of Roux-en-Y gastric bypass. The participant’s
et al (5 years; 120 participants)4 and Mingrone et al (10 years; postoperative course was complicated with alcohol use disorder and the
second liver biopsy 7 years after Roux-en-Y gastric bypass showed progression
60 participants).5 The study by Ikramuddin et al reported
to cirrhosis. The second patient had simple hepatic steatosis (without MASH
achievement of a composite outcome (HbA1c <7.0%, LDL and fibrosis) on the initial liver biopsy at the time of Roux-en-Y gastric bypass
<100 mg/dL, and systolic blood pressure <130 mm Hg) in 23% and the postoperative course was complicated with decompensated
of 57 patients who underwent a Roux-en-Y gastric bypass vs alcohol-associated cirrhosis, which was confirmed with a second liver biopsy
8 years after Roux-en-Y gastric bypass.
only 4% of 56 patients who underwent medical/lifestyle c
Excluding nonmelanoma skin cancer.
interventions. Likewise, the single-site RCT of Roux-en-Y
gastric bypass vs biliopancreatic diversion vs conventional Beyond glycemia, diabetes care goals extend to control of
medical treatment from Mingrone et al reported diabetes blood pressure and lipids and prevention of diabetes-related
remission at 10 years in 37.5% of 40 patients who underwent complications. Although no differences were observed be-
an operation compared with 5.5% of 18 patients in the medi- tween the groups in blood pressure or LDL cholesterol, bar-
cal group. Similar to ARMMS-T2D, these studies observed iatric surgery was superior at raising HDL and lowering triglyc-
a waning over time of surgical benefits on diabetes metrics. erides, which is consistent with previous smaller RCTs with

jama.com (Reprinted) JAMA February 27, 2024 Volume 331, Number 8 661

© 2024 American Medical Association. All rights reserved.

Downloaded from jamanetwork.com by College of Medicine of Seoul National Univ. user on 03/23/2024
 Research Original Investigation Long-Term Outcomes of Medical Management vs Bariatric Surgery in Type 2 Diabetes

shorter follow-up.31 Although differences in blood pressure Limitations

were not present, differential antihypertensive medication use This study has several limitations. First, all included trials were
between the groups may have confounded the results. A 2018 open-label. Second, there were differences in original trial pro-
RCT comparing surgical vs medical/lifestyle interventions dedi- tocols and lack of uniform assessment for some diabetes com-
cated to blood pressure as the primary outcome reported sur- plications, including retinopathy. Third, treatments exam-
gical superiority in this domain.32 ined in different trials were not identical. Randomization ratios
Traditionally, the minimum BMI threshold for use of bar- were not all 1:1 (because some trials had 2 surgical groups). Trial
iatric surgery among people with type 2 diabetes is 35.33 In an populations may have had different distributions of effect
exploratory analysis, reductions in HbA1c were similar among modifiers or outcome predictors, so a pooled analysis such as
participants randomized to the surgical group with baseline this could create bias, or a form of confounding induced by the
BMI at or above 35 or less than 35, and the difference in per- original trials. There may also have been selective dropout over
cent weight loss between the groups was greater in those with time in the medical/lifestyle group; for example, those who
BMI less than 35. Data from other RCTs have also shown that were doing better from a health perspective may have partici-
bariatric surgery can be effective for type 2 diabetes treat- pated longer. Fourth, there were different enrollment time-
ment for those with BMI less than 35.31 These data support the lines, resulting in different numbers of participants with dif-
newer standards for use of bariatric surgery among patients ferent lengths of follow-up.
with a lower BMI and type 2 diabetes, such as those recom- Fifth, the assumption that data are missing at random was
mended by the Diabetes Surgery Summit and joint interna- untestable with this study design. The per-protocol analysis
tional surgical societies.31,34 relied on strong assumptions about the relationship between
Anemia and fractures were more common after bariatric the (time-varying) treatment and the outcome in the model,
surgery. Micronutrient deficiencies may contribute to the the absence of uncontrolled time-varying confounding, and
higher fracture rate after bariatric surgery, and the potential the specification of the model for the inverse probability
for these and other deficiencies should be proactively moni- weight. Sixth, this analysis was not powered to detect differ-
tored and measured life-long in patients after undergoing bar- ences among the 3 surgical procedures for primary outcomes
iatric surgical procedures. Similar to other studies with shorter or to examine the risk of major adverse cardiovascular events,
follow-up,31 gastrointestinal adverse events were higher after microvascular events, cancer, or mortality. Seventh, bariatric
bariatric surgery. surgical procedures changed over the course of follow-up
Strengths of this analysis include inclusion of trials from (eg, reduction in use of adjustable gastric banding). Eighth, in-
4 different sites, which may increase generalizability com- creased use of incretin medications occurred in parallel in both
pared with the single-site composition of most previous RCTs.31 groups during follow-up. As a result, changes in use of this class
At baseline, more than one-third of participants had a BMI of of medication over time could have influenced between-
less than 35, whereas most previous studies of bariatric sur- group differences in outcomes. Ninth, consistent with other
gery focused on people with a BMI of 35 or greater. Three bar- bariatric surgery studies, most participants were women.
iatric surgical procedures were examined and data on sleeve
gastrectomy, which is now the most commonly performed bar-
iatric surgery operation worldwide, are included. More than
one-fourth of the study participants were of racial and ethnic
Conclusions
minority groups, which is relatively diverse compared with At 7 to 12 years of follow-up, individuals originally randomized
other studies of bariatric surgery.31 Additionally, the sample to undergo bariatric surgery, compared with medical/lifestyle in-
size and length of follow-up is substantially greater than in any tervention, had superior glycemic control with less diabetes
previous single RCT. medication usage and higher rates of diabetes remission.

ARTICLE INFORMATION (Vernon); Metamor Institute, Pennington Schauer, Kashyap, Cummings, Kirwan.
Accepted for Publication: January 10, 2024. Biomedical Research Center, Baton Rouge, Acquisition, analysis, or interpretation of data:
Louisiana (Schauer); Weill Cornell Medicine-New Courcoulas, Patti, Hu, Arterburn, Simonson,
Author Affiliations: Department of Surgery, York Presbyterian, Department of Medicine, Gourash, Vernon, Beck, Schauer, Kashyap, Aminian,
University of Pittsburgh, Pittsburgh, Pennsylvania Division of Endocrinology, Diabetes and Cummings, Kirwan.
(Courcoulas, Gourash); Research Division, Joslin Metabolism, New York, New York (Kashyap); Drafting of the manuscript: Courcoulas, Patti, Hu,
Diabetes Center, and Harvard Medical School, Bariatric and Metabolic Institute, Department of Simonson, Vernon, Kashyap, Cummings, Kirwan.
Boston, Massachusetts (Patti); Department of General Surgery, Cleveland Clinic, Cleveland, Ohio Critical review of the manuscript for important
Quantitative Health Sciences, Cleveland Clinic, (Aminian); Department of Medicine, University of intellectual content: Patti, Arterburn, Simonson,
Cleveland, Ohio (Hu, Beck); Kaiser Permanente Washington and VA Puget Sound Health Care Gourash, Jakicic, Vernon, Beck, Schauer, Kashyap,
Washington Health Research Institute, Seattle System, Seattle (Cummings); Pennington Aminian, Cummings, Kirwan.
(Arterburn); Division of Endocrinology, Diabetes Biomedical Research Center, Baton Rouge, Statistical analysis: Hu, Beck, Cummings, Kirwan.
and Hypertension, Brigham and Women’s Hospital, Louisiana (Kirwan). Obtained funding: Patti, Arterburn, Simonson,
and Harvard Medical School, Boston, Jakicic, Schauer, Kashyap, Cummings, Kirwan.
Massachusetts (Simonson); Department of Internal Author Contributions: Drs Courcoulas, Hu, and
Kirwan had full access to all of the data in the study Administrative, technical, or material support:
Medicine, Division of Physical Activity and Weight Courcoulas, Patti, Gourash, Jakicic, Beck, Schauer,
Management, University of Kansas Medical Center, and take responsibility for the integrity of the data
and the accuracy of the data analysis. Kashyap, Aminian, Cummings, Kirwan.
Kansas City (Jakicic); Division of General & GI Supervision: Patti, Simonson, Schauer, Kashyap,
Surgery, Brigham and Women’s Hospital, and Concept and design: Courcoulas, Patti, Hu,
Arterburn, Simonson, Gourash, Jakicic, Vernon, Cummings, Kirwan.
Harvard Medical School, Boston, Massachusetts

662 JAMA February 27, 2024 Volume 331, Number 8 (Reprinted) jama.com

© 2024 American Medical Association. All rights reserved.

Downloaded from jamanetwork.com by College of Medicine of Seoul National Univ. user on 03/23/2024
 Long-Term Outcomes of Medical Management vs Bariatric Surgery in Type 2 Diabetes Original Investigation Research

Conflict of Interest Disclosures: Dr Courcoulas during the original trial. All 4 studies above received a survey of policies. Obes Surg. 2020;30(2):707-713.
reported receiving grants from Alllurion and Eli Lilly bridge funding from Covidien (now Medtronic) and doi:10.1007/s11695-019-04243-2
outside the submitted work. Dr Patti reported Ethicon in preparation for the current NIH-funded 7. Aminian A, Nissen SE. Success (but unfinished)
receiving grants from National Institutes of Health observational follow-up and analysis. story of metabolic surgery. Diabetes Care. 2020;43
during the conduct of the study and grants from Role of the Funder/Sponsor: The sponsor (NIDDK) (6):1175-1177. doi:10.2337/dci20-0006
Dexcom; personal fees from Hanmi, MBX, and requested proposals for a randomized trial for
AstraZeneca; and serving on a data and safety 8. Simonson DC, Hu B, Arterburn DE, et al. Alliance
metabolic/bariatric surgery but had no role in the of Randomized Trials of Medicine vs Metabolic
monitoring board from Fractyl outside the study design of the original trials. Original trial
submitted work. Dr Hu reported receiving grants Surgery in Type 2 Diabetes (ARMMS-T2D): study
funding is detailed above. NIDDK was not involved rationale, design, and methods. Diabetes Obes Metab.
from NIH/NIDDK during the conduct of the study. in data collection but was involved in discussion of
Dr Arterburn reported receiving grants from NIDDK 2022;24(7):1206-1215. doi:10.1111/dom.14680
the analysis and interpretation of data for
during the conduct of the study and grants from ARMMS-T2D. An NIDDK project scientist was 9. Kirwan JP, Courcoulas AP, Cummings DE, et al.
NIH, PCORI, and Sharecare and nonfinancial included in the planning and editing of the report as Diabetes remission in the Alliance of Randomized
support from American Society of Metabolic and well as the decision to submit the manuscript for Trials of Medicine Versus Metabolic Surgery in
Bariatric Surgery for travel outside the submitted publication. Type 2 Diabetes (ARMMS-T2D). Diabetes Care.
work. Dr Simonson reported receiving grants from 2022;45(7):1574-1583. doi:10.2337/dc21-2441
NIH/NIDDK during the conduct of the study and Data Sharing Statement: See Supplement 3.
10. Schauer PR, Kashyap SR, Wolski K, et al.
being a stockholder/shareholder in GI Windows Additional Contributions: We wish to thank the Bariatric surgery versus intensive medical therapy
outside the submitted work. Dr Gourash reported National Institute of Diabetes and Digestive and in obese patients with diabetes. N Engl J Med. 2012;
receiving grants from NIH/NIDDK during the Kidney Diseases project scientists Karen Teff, PhD, 366(17):1567-1576. doi:10.1056/NEJMoa1200225
conduct of the study. Dr Jakicic reported receiving and Jean M. Lawrence, ScD. We also want to
personal fees from Wondr Health, Education express our gratitude to all of the study 11. Schauer PR, Bhatt DL, Kirwan JP, et al;
Initiatives, WW International, and Epitomee participants, surgeons, and clinical and research STAMPEDE Investigators. Bariatric surgery versus
Medical outside the submitted work. Dr Beck staff, as well as to Emily Eagleton, MS; Giovanna intensive medical therapy for diabetes–3-year
reported receiving grants from NIDDK and NHLBI Febbraro Bochicchio, MS (University of Pittsburgh); outcomes. N Engl J Med. 2014;370(21):2002-2013.
during the conduct of the study. Dr Schauer Kathleen Foster, BS; Danielle Wolfs, MPH (Joslin doi:10.1056/NEJMoa1401329
reported receiving grants from NIDDK/NIH during Diabetes Center); Janine Bauman, BSN; Chytaine 12. Courcoulas AP, Goodpaster BH, Eagleton JK,
the conduct of the study and personal fees from GI Hall, AAN (Cleveland Clinic); Reba Blissell, RN; and et al. Surgical vs medical treatments for type 2
Dynamics, Persona, Mediflix, Metabolic Health Katie Wicklander, BS (University of Washington), for diabetes mellitus: a randomized clinical trial. JAMA
Institute, Lilly, SE Healthcare, lder, grants from study coordination; Christopher Axelrod, MSc Surg. 2014;149(7):707-715. doi:10.1001/jamasurg.
Ethicon, personal fees from Ethicon Honoraria for (Pennington Biomedical Research Center), for 2014.467
speaking, grants from Medtronic, personal fees technical assistance; the investigators of the 13. Courcoulas AP, Belle SH, Neiberg RH, et al.
from Medtronic Honoraria for speaking, personal original trials; and the study volunteers for the Three-year outcomes of bariatric surgery vs lifestyle
fees from Novo Nordisk Honoraria for speaking, and considerable time and effort on the trial. The intervention for type 2 diabetes mellitus treatment:
personal fees from Heron Advisory Board outside authors also thank the Cleveland Clinic Clinical a randomized clinical trial. JAMA Surg. 2015;150
the submitted work. Dr Kashyap reported receiving Research Unit and the Joslin Clinical Research (10):931-940. doi:10.1001/jamasurg.2015.1534
nonfinancial support from Fractyl Laboratories, Center for technical support. None of the named
personal fees from GI Dynamics, and serving as individuals were compensated for their 14. Courcoulas AP, Gallagher JW, Neiberg RH, et al.
contractual chief medical officer for Gila contributions. Bariatric surgery vs lifestyle intervention for
Therapeutics outside the submitted work. diabetes treatment: 5-year outcomes from a
Dr Aminian reported receiving grants and personal REFERENCES randomized trial. J Clin Endocrinol Metab. 2020;105
fees from Medtronic, Eli Lilly, and Ethicon outside (3):866-876. doi:10.1210/clinem/dgaa006
1. Sun H, Saeedi P, Karuranga S, et al. IDF Diabetes
the submitted work. Dr Cummings reported serving Atlas: global, regional and country-level diabetes 15. Halperin F, Ding SA, Simonson DC, et al.
on scientific advisory boards of GI Dynamics, prevalence estimates for 2021 and projections for Roux-en-Y gastric bypass surgery or lifestyle with
Endogenex, and Gila Pharmaceuticals. Dr Kirwan 2045. Diabetes Res Clin Pract. 2022;183:109119. intensive medical management in patients with
reported grants from NIH/NIDDK U01 DK114156 doi:10.1016/j.diabres.2021.109119 type 2 diabetes: feasibility and 1-year results of a
during the conduct of the study. No other randomized clinical trial. JAMA Surg. 2014;149(7):
disclosures were reported. 2. American Diabetes Association. Economic costs 716-726. doi:10.1001/jamasurg.2014.514
of diabetes in the US in 2017. Diabetes Care. 2018;
Funding/Support: ARMMS-T2D is supported by 41(5):917-928. doi:10.2337/dci18-0007 16. Ding SA, Simonson DC, Wewalka M, et al.
cooperative agreement U01 DK114156 from the Adjustable gastric band surgery or medical
National Institute of Diabetes and Digestive and 3. Schauer PR, Bhatt DL, Kirwan JP, et al; management in patients with type 2 diabetes:
Kidney Diseases (NIDDK). Funding for the original STAMPEDE Investigators. Bariatric surgery versus a randomized clinical trial. J Clin Endocrinol Metab.
trials were as follows: STAMPEDE was supported by intensive medical therapy for diabetes: 5-year 2015;100(7):2546-2556. doi:10.1210/jc.2015-1443
an investigator-initiated grant from Ethicon outcomes. N Engl J Med. 2017;376(7):641-651. doi:
10.1056/NEJMoa1600869 17. Simonson DC, Halperin F, Foster K, Vernon A,
Endo-Surgery and in-kind support from LifeScan Goldfine AB. Clinical and patient-centered
and NIDDK grant R01 DK089547. TRIABETES was 4. Ikramuddin S, Korner J, Lee WJ, et al. Lifestyle outcomes in obese patients with type 2 diabetes
funded by NIDDK grants RC1DK086037 and R01 intervention and medical management with vs 3 years after randomization to Roux-en-Y gastric
DK095128, and by subsidization of the surgical without Roux-en-Y gastric bypass and control of bypass surgery versus intensive lifestyle
procedures by Magee-Womens Hospital of the hemoglobin A1c, LDL cholesterol, and systolic blood management: the SLIMM-T2D study. Diabetes Care.
University of Pittsburgh Medical Center. pressure at 5 years in the Diabetes Surgery Study. 2018;41(4):670-679. doi:10.2337/dc17-0487
CROSSROADS was funded by NIDDK grant R01 JAMA. 2018;319(3):266-278. doi:10.1001/jama.2017.
DK089528, as well as by grants from the Group 20813 18. Simonson DC, Vernon A, Foster K, Halperin F,
Health Research Institute and Group Health Patti ME, Goldfine AB. Adjustable gastric band
5. Mingrone G, Panunzi S, De Gaetano A, et al. surgery or medical management in patients with
Foundation. The Shared Decision-Making aid used Metabolic surgery versus conventional medical
in this trial was provided by the Informed Medical type 2 diabetes and obesity: three-year results of a
therapy in patients with type 2 diabetes: 10-year randomized trial. Surg Obes Relat Dis. 2019;15(12):
Decisions Foundation. SLIMM-T2D was funded by follow-up of an open-label, single-centre,
NIDDK grants RC1 DK086918, R56 DK095451, and 2052-2059. doi:10.1016/j.soard.2019.03.038
randomised controlled trial. Lancet. 2021;397
P30 DK036836; the Herbert Graetz Fund at Joslin (10271):293-304. doi:10.1016/S0140-6736(20) 19. Cummings DE, Arterburn DE, Westbrook EO,
Diabetes Center; and Patient-Centered Outcomes 32649-0 et al. Gastric bypass surgery vs intensive lifestyle
Research Institute grant CE-1304-6756. LifeScan, and medical intervention for type 2 diabetes: the
a division of Johnson & Johnson, provided home 6. Gebran SG, Knighton B, Ngaage LM, et al. CROSSROADS randomised controlled trial.
glucose-monitoring supplies; Nestle provided Insurance coverage criteria for bariatric surgery: Diabetologia. 2016;59(5):945-953. doi:10.1007/
Boost; and Novo Nordisk provided drug supplies s00125-016-3903-x

jama.com (Reprinted) JAMA February 27, 2024 Volume 331, Number 8 663

© 2024 American Medical Association. All rights reserved.

Downloaded from jamanetwork.com by College of Medicine of Seoul National Univ. user on 03/23/2024
 Research Original Investigation Long-Term Outcomes of Medical Management vs Bariatric Surgery in Type 2 Diabetes

20. Arterburn D, Flum DR, Westbrook EO, et al; 26. Lean MEJ, Leslie WS, Barnes AC, et al. evidence for the legacy effect of surgery. Diabetes
CROSSROADS Study Team. A population-based, Durability of a primary care-led weight- Care. 2016;39(8):1400-1407. doi:10.2337/dc16-0194
shared decision-making approach to recruit for a management intervention for remission of type 2 31. Rubino F, Nathan DM, Eckel RH, et al; Delegates
randomized trial of bariatric surgery versus lifestyle diabetes: 2-year results of the DiRECT open-label, of the 2nd Diabetes Surgery Summit. Metabolic
for type 2 diabetes. Surg Obes Relat Dis. 2013;9(6): cluster-randomised trial. Lancet Diabetes Endocrinol. surgery in the treatment algorithm for type 2
837-844. doi:10.1016/j.soard.2013.05.006 2019;7(5):344-355. doi:10.1016/S2213-8587(19) diabetes: a joint statement by International
21. Buse JB, Caprio S, Cefalu WT, et al. How do we 30068-3 Diabetes Organizations. Diabetes Care. 2016;39(6):
define cure of diabetes? Diabetes Care. 2009;32 27. Knowler WC, Barrett-Connor E, Fowler SE, et al; 861-877. doi:10.2337/dc16-0236
(11):2133-2135. doi:10.2337/dc09-9036 Diabetes Prevention Program Research Group. 32. Schiavon CA, Bersch-Ferreira AC, Santucci EV,
22. Riddle MC, Cefalu WT, Evans PH, et al. Reduction in the incidence of type 2 diabetes with et al. Effects of bariatric surgery in obese patients
Consensus report: definition and interpretation of lifestyle intervention or metformin. N Engl J Med. with hypertension: the GATEWAY randomized trial
remission in type 2 diabetes. Diabetes Care. 2021; 2002;346(6):393-403. doi:10.1056/NEJMoa012512 (Gastric Bypass to Treat Obese Patients With
44(10):2438-2444. doi:10.2337/dci21-0034 28. Wing RR, Bolin P, Brancati FL, et al; Look Steady Hypertension). Circulation. 2018;137(11):
23. van der Wal WM, Geskus RB. ipw: an R package AHEAD Research Group. Cardiovascular effects of 1132-1142. doi:10.1161/CIRCULATIONAHA.117.032130
for inverse probability weighting. J Stat Softw. 2011; intensive lifestyle intervention in type 2 diabetes. 33. Consensus Development Conference Panel.
43(13):1-23. N Engl J Med. 2013;369(2):145-154. doi:10.1056/ NIH conference: gastrointestinal surgery for severe
NEJMoa1212914 obesity. Ann Intern Med. 1991;115(12):956-961.
24. Austin PC, Stuart EA. Moving towards best
practice when using inverse probability of 29. Lean ME, Leslie WS, Barnes AC, et al. Primary doi:10.7326/0003-4819-115-12-956
treatment weighting (IPTW) using the propensity care-led weight management for remission of 34. Eisenberg D, Shikora SA, Aarts E, et al. 2022
score to estimate causal treatment effects in type 2 diabetes (DiRECT): an open-label, American Society for Metabolic and Bariatric
observational studies. Stat Med. 2015;34(28):3661- cluster-randomised trial. Lancet. 2018;391(10120): Surgery (ASMBS) and International Federation for
3679. doi:10.1002/sim.6607 541-551. doi:10.1016/S0140-6736(17)33102-1 the Surgery of Obesity and Metabolic Disorders
25. Hernán MA, Brumback BA, Robins JM. 30. Coleman KJ, Haneuse S, Johnson E, et al. (IFSO): indications for metabolic and bariatric
Estimating the causal effect of zidovudine on CD4 Long-term microvascular disease outcomes in surgery. Surg Obes Relat Dis. 2022;18(12):1345-1356.
count with a marginal structural model for repeated patients with type 2 diabetes after bariatric surgery: doi:10.1016/j.soard.2022.08.013
measures. Stat Med. 2002;21(12):1689-1709.
doi:10.1002/sim.1144

664 JAMA February 27, 2024 Volume 331, Number 8 (Reprinted) jama.com

© 2024 American Medical Association. All rights reserved.

Downloaded from jamanetwork.com by College of Medicine of Seoul National Univ. user on 03/23/2024