Results and Summary of Recommendations
A summary of all recommendations is provided in [Table 1].
In adults with overweight or obesity, the ASGE–ESGE suggests the use of EBMTs plus
LM over LM alone for patients with a body mass index (BMI) of ≥ 30 kg/m2 or BMI of 27.0 to 29.9 kg/m2 with at least 1 obesity-related comorbidity.
(Conditional recommendation, very low certainty)
Implementation considerations
-
For patients with a BMI of 27.0 to 29.9 kg/m2 with at least 1 obesity-related comorbidity, data were available for IGB, EGR, and
DJBL.
-
For patients with class III obesity, data were available for IGB, EGR, AT, and DJBL.
Summary of the evidence
For the subgroup with BMIs of 27.0 to 29.9 kg/m2, 6 observational studies were used to inform this PICO (IGB studies [55]
[56], EGR study [57], and DJBL studies [58]
[60]). Of these, 6 studies were used to assess safety [55]
[56]
[57]
[58]
[59]
[60], 4 studies for percentage of TWL [55]
[56]
[57]
[59], and 3 studies for the change in HbA1c [58]
[59]
[60]. All studies on IGB and EGR only included patients who were overweight (BMI of 25.0–29.9 kg/m2 or 27.0–29.9 kg/m2). All DJBL studies included patients who were both overweight (starting BMI of 27.0
or 28.0 kg/m2) and had obesity. Mean age ranged from 38 to 52 years and BMI from 29.7 to 43.1 kg/m2. All studies prescribed concomitant LM, except for Moore et al [56], where the intensity of LM varied across sites given the nature of real-world experience
(Supplementary Table 2, available online).
For the subgroup with classes I and II obesity, 17 RCTs were used to inform this PICO
(IGB studies [39]
[40]
[41]
[42]
[43]
[44], EGR studies [45]
[46]
[47]
[62], AT studies [48]
[52], TPS studies [49], DJBL studies [50]
[51]
[63], and DMR studies [64]). Of these, 15 studies were used to assess safety [39]
[40]
[42]
[43]
[44]
[45]
[46]
[49]
[50]
[51]
[52]
[61]
[62]
[63]
[64],14 studies for percentage of TWL [39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[49]
[50]
[51]
[52]
[64], and 2 studies for the change in HbA1c [50]
[51]. All IGB, EGR, and TPS studies only included patients with classes I and II obesity.
Otherwise, the remainder included a combination of different classes of obesity (classes
II and III for AT; classes I, II, and III for DJBL; and overweight and classes I and
II for DMR). Mean age ranged from 38 to 58 years and BMI from 31.5 to 42.0 kg/m2. Most studies compared EBMTs with LM alone, whereas Sullivan et al [44], Ponce et al [43], Sullivan et al [47], Rothstein et al [49], Thompson et al [50], and Mingrone et al [64] compared EBMTs with sham (Supplementary Table 3, available online).
For the subgroup with class III obesity, 31 observational studies and RCTs (interventional
arms only) were used to inform this PICO (IGB studies [55]
[56]
[61]
[65]
[66]
[67]
[68]
[69]
[70]
[71]
[72]
[73], EGR studies [57]
[74], AT studies [48]
[52]
[75], and DJBL studies [50]
[51]
[58]
[59]
[63]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84]). Of these, 26 studies were used to assess safety [48]
[50]
[51]
[52]
[57]
[58]
[59]
[63]
[67]
[68]
[69]
[70]
[71]
[72]
[73]
[74]
[75]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84], 20 studies for percentage of TWL [48]
[50]
[51]
[52]
[55]
[56]
[57]
[59]
[61]
[65]
[66]
[67]
[68]
[69]
[71]
[72]
[73]
[74]
[75]
[81], and 10 studies for the changes in HbA1c [50]
[51]
[58]
[59]
[79]
[80]
[81]
[82]
[83]
[84]. All IGB and EGR studies only included patients with class III obesity, whereas
AT and DJBL studies included both class III and other classes of obesity (class II
± class I). Mean age ranged from 33 to 58 years and BMI from 34.6 to 69.1 kg/m2. All studies prescribed concomitant LM, except for Moore et al [56], where the intensity of LM varied across sites given the nature of real-world experience
(Supplementary Table 4, available online).
Benefits
For the subgroup with BMIs of 27.0 to 29.9 kg/m2, 4 observational studies (n = 692) informed the outcomes of percentage of TWL at
6 months (for IGB) or 12 months (for EGR and DJBL) and 3 studies (n = 436) for HbA1c
reduction at 12 months (for DJBL) [55]
[56]
[57]
[58]
[59]
[60]. The pooled weight loss at 6 to 12 months was 11.9 % TWL (95 % confidence interval
[CI], 7.7–16.0) (Supplementary Fig. 1, available online) and pooled HbA1c reduction at 12 months was 1.0 % (95 % CI, .6–1.5)
(Supplementary Fig. 2, available online).
For the subgroup with classes I and II obesity, 14 RCTs (n = 2787) informed the outcomes
of percentage of TWL at 12 months [39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[49]
[50]
[51]
[52]
[64] and 2 studies (n = 490) for HbA1c reduction at 12 months [50]
[51]. A total of 1636 subjects were in the EBMT plus LM group and 1151 in the LM group.
The mean difference (MD), which represented the difference between the pooled percentage
of TWL in the EBMT arm minus the control arm, at IGB removal or 12 months after EGR,
AT, DJBL, or TPS was 7.1 % (95 % CI, 5.4–8.8) in favor of EBMT (Supplementary Fig. 3, available online). The mean TWL of the EBMT arm ranged from 5.0 % to 18.6 % at IGB
removal or 12 months for EGR, DJBL, AT, or TPS. The MD, which represented the difference
between the pooled HbA1c reduction in the EBMT arm minus the control arm, at 12 months
was .7 % (95 % CI, .4–1.1) in favor of EBMT (Supplementary Fig. 4, available online). The mean HbA1c reduction of the EBMT arms ranged from 1.1 % to
1.5 % at 12 months.
For the subgroup with class III obesity, 20 observational studies (n = 2776) informed
the outcomes of percentage of TWL at 6 to 12 months [48]
[50]
[51]
[52]
[55]
[56]
[57]
[59]
[61]
[65]
[66]
[67]
[68]
[69]
[71]
[72]
[73]
[74]
[75]
[81] and 10 studies (n = 815) for HbA1c reduction at 12 months [50]
[51]
[58]
[59]
[79]
[80]
[81]
[82]
[83]
[84]. The pooled TWL at 6 to 12 months was 13.1 % (95 % CI, 10.8–15.4) (Supplementary Fig. 5, available online) and pooled HbA1c reduction at 12 months was 1.3 % (95 % CI, 1,0–1.6)
(Supplementary Fig. 6, available online).
Harms
For the subgroup with BMIs of 27.0 to 29.9 kg/m2, 6 observational studies informed the outcome of serious adverse events (SAEs; n = 7416)
[55]
[56]
[57]
[58]
[59]
[60]. SAEs were defined by the investigators and reported in the original studies. The
pooled estimate for SAEs showed an event rate of 2.7 % (95 % CI, 1.2–6.0) (Supplementary Fig. 7, available online).
For the subgroup with classes I and II obesity, 16 RCTs informed the outcome of SAEs
(n = 1464) [39]
[40]
[42]
[43]
[44]
[45]
[46]
[49]
[50]
[51]
[52]
[62]
[63]
[64]. The pooled estimate for SAEs showed an absolute risk of 14 additional SAEs per
1000 subjects [6]
[30] in the EBMT group (111/2135) compared with the control group (6/1464) (Supplementary Fig. 8, available online).
For the subgroup with class III obesity, 26 studies informed the outcome of SAEs (n = 2042)
[48]
[50]
[51]
[52]
[57]
[58]
[59]
[63]
[67]
[68]
[69]
[70]
[71]
[72]
[73]
[74]
[75]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84].The pooled estimate for SAEs showed an event rate of 6.9 % (95 % CI, 5.7–8.2) (Supplementary Fig. 9, available online).
Certainty of evidence assessment
The certainty in the evidence of effects of EBMTs in the subgroup with BMIs of 27.0
to 29.9 kg/m2 with at least 1 comorbidity, subgroup with classes I to II obesity, and subgroup
with class III obesity was very low, low, and very low, respectively (Supplementary Table 5, available online). Therefore, the overall certainty in the evidence of this PICO
(ie, the effects of EBMTs for patients with a BMI of ≥ 30 kg/m2 or 27.0–29.9 kg/m2 with ≥ 1 comorbidity) was deemed to be very low.
In the subgroup with BMI of 27.0 to 29.9 kg/m2, for the weight loss outcome, there was a concern for confounding bias in some studies
as well as inconsistency and indirectness because some studies reported the amount
of weight loss in both the overweight and obesity groups combined. For the HbA1c outcome,
there was a concern for inconsistency, indirectness (because of a mixed overweight
and obesity population in some studies), and imprecision (because of a small total
number of patients). For harms, there was a very low certainty in evidence given the
inconsistency, indirectness (because of a mixed overweight and obesity population
in some studies), and imprecision (because of a small number of SAEs) (Supplementary Table 6, available online).
In the subgroup with classes I and II obesity, there was inconsistency in the amount
of weight loss, which was likely explained by the heterogeneity among different EBMT
devices and/or procedures pooled. For the HbA1c outcome, there was imprecision because
the CI crossed the line of no difference. For harms, the certainty of evidence was
downgraded twice for imprecision because of a low event rate and wide CI (Supplementary Table 7, available online).
In the subgroup with class III obesity, for the weight loss outcome, there was a concern
for confounding bias in some studies as well as inconsistency and indirectness because
some studies reported the amount of weight loss of both class III obesity and other
classes combined. For the HbA1c outcome, there was a concern for inconsistency and
indirectness with some studies reporting the outcomes of both class III obesity and
other classes combined. For harms, there was a very low certainty in evidence given
the inconsistency, indirectness, and imprecision because of a small number of SAEs
(Supplementary Table 8, available online).
Discussion
To assess the patient populations in which EBMTs should be considered, we divided
the potential populations into 3 categories based on BMI: BMI of 27.0 to 29.9 kg/m2 with at least 1 obesity-related comorbidity, classes I and II obesity, and class
III obesity. Because most EBMTs included in this guideline were approved or cleared
for classes I and II obesity, only RCTs were included for this population. In contrast,
for the BMI of 27.0 to 29.9 kg/m2 and class III obesity subgroups, no RCTs specifically assessed the effect of EBMTs
in these 2 populations. Therefore, observational studies were evaluated.
For the overweight category, whereas Moore et al [56] and Barrichello et al [57] included patients with BMIs of 25.0 to 29.9 kg/m2, most studies included patients starting at BMIs of 27 or 28 kg/m2
[55]
[56]
[57]
[58]
[59]
[60]. Additionally, half of the studies included patients with at least 1 obesity-related
comorbidity (T2DM). Therefore, the panel decided to use a conservative cutoff for
this patient population with a starting BMI of 27 kg/m2 with at least 1 comorbidity. For the class III obesity category, all IGB and EGR
studies [55]
[56]
[57]
[61]
[65]
[66]
[67]
[68]
[69]
[70]
[71]
[72]
[73]
[74] only included patients with class III obesity, whereas AT and DJBL studies [48]
[50]
[51]
[52]
[58]
[59]
[75]
[79]
[80]
[81]
[82]
[83]
[84] included both class III and class II ± class I obesity. Although some studies had
a cutoff for the highest BMI at 50 or 55 kg/m2
[48]
[50]
[51]
[52]
[55]
[56]
[57]
[58]
[59]
[61]
[63]
[69]
[70]
[74]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84], some did not and recruited patients with BMIs up to 70 or 78 kg/m2
[65]
[66]
[67]
[68]
[71]
[72]
[73]
[75]
[82]. The panel accepted the heterogeneity in this patient population. However, given
that EBMTs may be used for either primary therapy or bridge therapy before bariatric
surgery, the panel agreed to not having an upper limit of BMI for consideration of
EBMTs.
The amount of weight loss after EBMT was determined to be moderate for all BMI subgroups.
Specifically, the amount of weight loss was 11.9 % (95 % CI, 7.7–16.0) and 13.1 %
(95 % CI, 10.8–15.4) TWL in the BMI of 27.0 to 29.9 kg/m2 and class III obesity subgroups, respectively. For the subgroup with classes I and
II obesity, the MD, representing the difference between the pooled percentage of TWL
in the EBMT arm minus the control arm, was 6.3 % (95 % CI, 5.3–7.3) in favor of EBMT,
with the absolute percentage of TWL in the EBMT arm ranging from 5.0 % to 18.6 % at
12 months. For the BMI of 27.0 to 29.9 kg/m2 and class III obesity studies, the lower CI of percentage of TWL was 7.7 % and 10.8 %,
respectively. Given the pooled average of 3.2 % TWL for the historical control subjects
from all EBMT RCTs (Supplementary Fig. 10, available online), the MD of the amount of weight loss between the EBMT and control
groups in these 2 populations should remain above the 3 % TWL minimal important difference
threshold (MDs of 4.5 % and 7.6 % TWL, respectively). Similarly, for the subgroup
with classes I and II obesity, not only did the lower CI of the overall MD lie above
the 3 % TWL minimal important difference threshold, but our sensitivity analysis also
showed that the lower CI of the MD of every EBMT also lay above this threshold (Supplementary Fig. 3). Additionally, all studies but IGB reported the weight loss outcome at 12 months.
For IGB, all studies reported percentage of TWL at the time of IGB removal (6–8 months).
Although Nunes et al [85] reported percentage of TWL in the subgroups with overweight and class III obesity
at 12 months (ie, 6 months after IGB removal), this study evaluated the effect of
IGB plus a very-low-calorie diet, which likely biased the magnitude of weight loss
[86]. Therefore, this study was excluded. The effect of IGB on weight loss after IGB
removal in the subgroups with overweight and class III obesity therefore remains to
be assessed. The panel also noted inconsistency in the amount of weight loss, especially
for class III obesity. This was believed to be because of a heterogeneity of the patient
populations, with some studies including patients with BMIs up to 55 kg/m2 for a primary therapy as an alternative to bariatric surgery [48]
[50]
[51]
[52]
[55]
[56]
[57]
[58]
[59]
[61]
[63]
[69]
[70]
[74]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84] and others including patients with BMIs up to 78 kg/m2 for bridge therapy before bariatric surgery [65]
[66]
[67]
[68]
[71]
[72]
[73]
[75]
[82]. The certainty of evidence was downgraded because of this inconsistency.
The SAE rate was 2.7 % (95 % CI, 1.2–6.0) and 6.9 % (95 % CI, 5.7–8.2) for the BMI
of 27.0 to 29.9 kg/m2 and class III obesity subgroups, respectively. For the subgroup with classes I and
II obesity, the risk ratio of SAEs in the EBMT arm compared with the control arm was
4.4 (95 % CI, 2.4–8.2), which was equivalent to 14 additional events per 1000 subjects.
The SAE rate in the EBMT arm ranged from 0 % to 10.6 %. Of note, the panel found that
the wide CIs for pooled SAE rates were likely because of the difference in SAE definitions
used by the authors, especially for DJBL studies. For example, although most DJBL
studies defined SAEs as those resulting in early device explantation, Stratmann et
al [82] only reported the rate of early device explantation and Roehlen et al [77] only reported the rate of SAEs without reporting the number of early device explantations.
In contrast, early removal of IGBs has not been considered as a SAE in most trials,
and specifically in the United States, RCTs would not meet the FDA categorization
of SAE by itself.
Currently, the number of studies evaluating the cost-effectiveness of EBMTs is limited.
Saumoy et al [87] and Kelly et al [88] demonstrated that ESG was cost-effective compared with LM alone in class II obesity
in the United States and United Kingdom, respectively. Haseeb et al [89] showed that ESG was cost-effective compared with GLP-1RA and sleeve gastrectomy
in class II obesity in the United States. Although currently no study has specifically
evaluated the cost-effectiveness of EBMTs in other obesity classes or in an overweight
population, the panel agreed that EBMTs would most likely be cost-effective, especially
when compared with LM, in these other BMI categories.
The panel considered the current state of EBMTs to be associated with reduced equity
for all BMI subgroups. This is solely because of the lack of insurance coverage for
EBMTs in most countries. This leads to inequity between those patients who are able
to afford the procedures and those who are not and potentially between the nonminority
and minority. The panel noted that with universal insurance coverage, EBMTs will improve
equity by providing better access to safe and effective care for more patients who
suffer from obesity or overweight with at least 1 obesity-related comorbidity.
In adults with obesity, the ASGE–ESGE suggests the use of an IGB plus LM over LM alone.
(Conditional recommendation, moderate certainty)
Rationale
A conditional recommendation is driven primarily by moderate variability in patient
values and preferences. Specifically, although the IGB is generally acceptable among
most patients suffering from obesity, some may prefer a less-invasive treatment approach
(ie, LM) despite a lower weight loss than seen with the IGB. Therefore, treatment
options should be discussed to encourage shared decision-making.
Summary of the evidence
We identified a recently published guideline on IGB, which conducted a systematic
review and meta-analysis with a comprehensive search strategy (MEDLINE, Embase, and
Cochrane Library) from inception to January 2020 [36]
[90]. We updated the search to March 2021 and found no additional RCTs that met our inclusion
and exclusion criteria. Therefore, 7 RCTs assessing the safety and efficacy of IGB
were used to inform this PICO [39]
[40]
[41]
[42]
[43]
[44]
[91]. All 7 studies reported percentage of TWL at the time of IGB removal (6–8 months),
and 2 studies reported percentage of TWL at 12 months [40]
[41]. Mean age and BMI of the intervention arm ranged from 38.7 to 44.4 years and from
30.3 to 53.9 kg/m2, respectively. The interventional arm of all studies underwent concomitant LM. The
control arms of Sullivan et al [44] and Ponce et al [43] underwent a sham procedure with concomitant LM, whereas the rest of the studies
underwent LM alone (Supplementary Table 9, available online).
Benefits
Seven RCTs informed the outcome of percentage of TWL at the time of IGB removal (6–8
months) [39]
[40]
[41]
[42]
[43]
[44]
[91], and 2 RCTs informed the outcome of percentage of TWL at 12 months [40]
[41]. Seven hundred seventy-nine subjects were in the IGB plus LM group and 654 in the
LM group. The MD, representing the difference between the pooled percentage of TWL
in the IGB arm minus the control arm at the time of IGB removal (6–8 months), was
6.9 % TWL (95 % CI, 4.1–9.7) in favor of the intervention (Supplementary Fig. 11, available online). This represented a 3.1 times greater weight loss in the IGB arm
compared with the control arm (pooled weight loss of 10.7 % TWL in the IGB arm vs
3.4 % TWL in the control arm). The MD for percentage of TWL at 12 months was 4.4 %
TWL (95 % CI, 2.9–6.0) (Supplementary Fig. 12). This represented a 2.4 times greater weight loss in the IGB arm compared with the
control arm (pooled weight loss of 7.9 % TWL in the IGB arm vs 3.3 % TWL in the control
arm).
Harms
Seven RCTs informed the outcome of SAEs [39]
[40]
[41]
[42]
[43]
[44]
[91]. SAEs were defined by the investigators and reported in the original studies. The
pooled estimate for SAEs showed an absolute risk of 32 additional SAEs per 1000 subjects
(95 % CI, 7–114) in the IGB group (58/1028) compared with the control group (0/798)
(Supplementary Fig. 13, available online). Selected examples of SAEs from studies that reported particular
SAE outcomes included esophageal mucosal injury (4/473), gastric ulcer/bleeding (5/650),
severe dehydration (5/704), aspiration pneumonia (2/42), perforation (2/653), gastric
outlet/bowel obstruction (1/802), and mortality (0/741) (Supplementary Table 10, available online).
Certainty of evidence assessment
The overall certainty in the evidence of effects for IGB was moderate (Supplementary Tables 11 and 12 and Supplementary Fig. 14, available online). For benefits at 6 months, we found imprecision with weight loss
because of the wide CI and some inconsistency that was not deemed of serious concern
by itself, and no additional downgrading was performed. For benefits at 12 months,
imprecision was found because of a small sample size and CI that crossed the line
of no difference. For harms, there was moderate certainty in evidence given a small
number of SAEs with a wide CI.
Discussion
The first IGB approved for use was the Garren-Edwards Gastric Bubble (American Edwards
Laboratories, Irvine, Calif, USA) in 1985, an air-filled balloon made of polyurethane
in a cylindrical shape that was removed from the market in 1988 because of SAEs and
lack of effective weight loss [92]
[93]
[94]. Current IGBs have been designed to mitigate AEs and have demonstrated weight loss
efficacy in sham-controlled trials as noted in the summary of evidence. The next generation
of IGBs approved in the United States and Europe came in 2015 and 2017, respectively,
but IGBs have been used around the world since the 1990 s.
The mechanism of action of IGBs for weight loss is likely multifactorial. Early data
suggested that at least 400 mL of space occupation in the stomach was required to
reduce meal volume [95]. Subsequent analysis of gastric emptying has demonstrated that the effects of fluid-filled
IGBs are also in part because of a reduction in the rate of gastric emptying during
balloon implantation [96]. These mechanisms may help explain the recurrent weight gain that can occur after
balloon removal, because the currently understood mechanisms for weight loss require
balloon presence.
The magnitude of weight loss with IGB at 6 months was determined to be moderate, with
a wide CI based on the mix of sham-controlled and open-label RCTs included in the
analysis. An analysis comparing open-label and sham IGB RCTs found that the sham study
design lowered weight loss compared with open-label studies [97]. Combining open-label and sham-controlled studies in this analysis may underestimate
the true effect of IGB in a clinical setting; however, this is the most conservative
approach. Additionally, the panel noted that weight loss was lower at 12 months (6
months after IGB removal) than at IGB removal. Although weight loss at the 12-month
time point was still significant, patients considering IGB therapy should be made
aware of the likely regain of some weight within 6 months of IGB removal. Studies
have evaluated repeated use of IGB for longer term obesity treatment [98]
[99], but repeated IGB therapy was not evaluated in this recommendation.
SAEs were also discussed by the panel. The SAE rate was 5.6 %, but safety varied across
the gas-filled compared with fluid-filled balloons [39]
[40]
[43]
[44]. Of note, most SAEs were related to short-term accommodative symptoms including
nausea and vomiting, leading to dehydration and abdominal pain. Although these did
meet the FDA criteria for SAEs, they were short-lived and resolved without sequelae,
leading the panel to determine the reported rates of SAEs were acceptable.
The panel also found current reduced equity related to IGB treatment. This is solely
because of the lack of insurance coverage of IGB in most countries. This leads to
inequity between those patients who are able to pay out of pocket for IGB treatment
and those patients who are not. The panel noted that insurance coverage is crucial
to reduce inequity and improve access to recommended obesity treatments. The panel
found that acceptability of IGBs was high with the caveat of some recurrent weight
gain 6 months after IGB removal and noted that some patients favor the shorter duration
of treatment with no permanent changes to the anatomy of the GI tract.
In adults undergoing IGB placement, the ASGE–ESGE suggests the use of antiemetics
periprocedurally.
(Conditional recommendation, very low certainty)
Further details regarding the rationale for this recommendation including the results
of systematic reviews, expert survey, and evidence profile are presented in Appendix
2 (available online).
In adults undergoing IGB placement, the ASGE–ESGE suggests the use of pain medications
periprocedurally.
(Conditional recommendation, very low certainty)
Further details regarding the rationale for this recommendation including the results
of systematic reviews, expert survey, and evidence profile are presented in Appendix
3 (available online).
In adults undergoing IGB placement, the ASGE–ESGE suggests the use of proton pump
inhibitors (PPIs) while the IGB is in place over no PPIs.
(Conditional recommendation, very low certainty)
Further details regarding the rationale for this recommendation including the results
of systematic reviews, expert survey, and evidence profile are presented in Appendix
4 (available online).
In adults with obesity, the ASGE–ESGE suggests treatment with EGR plus LM over LM
alone.
(Conditional recommendation, moderate certainty)
Implementation consideration
-
EGR may be performed using the Overstitch Endoscopic Suturing System (Apollo Endosurgery),
Incisionless Operating Platform (IOP; USGI Medical), or Endomina System (Endo Tools
Therapeutics). Prolene sutures are placed in the stomach to reduce its volume in all
cases. The procedures have been generally referred to as endoscopic gastric plication
or ESG, originally described with the Overstitch Endoscopic Suturing System. The primary
obesity surgery endoluminal (POSE) procedure specifically referred to a procedure
with the IOP; however, these also have been referred to as plication ESG in the literature.
Evidence is insufficient to specifically recommend 1 device over another. The choice
of device is based on clinical context, patient values, availability, and operator
experience.
Rationale
A conditional recommendation is driven primarily by moderate variability in patient
values and preferences. Specifically, although EGR is generally acceptable among most
patients suffering from obesity, some may prefer a less-invasive treatment approach
(ie, LM) despite lower weight loss than seen with the EGR. Therefore, treatment options
should be discussed to encourage shared decision-making. In addition, insurance coverage
is frequently lacking. A greater number of patients would elect to get EGR if it were
universally covered by insurance. Furthermore, insurance coverage would reduce healthcare
inequity.
Summary of the evidence
Four RCTs assessing the safety and efficacy of EGR were used to inform this PICO [45]
[46]
[47]
[62]. Of these, 4 studies were used to assess safety [45]
[46]
[47]
[62], and 3 studies were used to assess efficacy [45]
[46]
[47]
[62]. In Huberty et al [62], the control arm was offered a crossover to the intervention arm at 6 months; therefore,
the efficacy, which is the difference in mean weight loss between 2 two arms at 12
months, was not able to be assessed. Of the 4 studies, 1 study [45] used the Overstitch suturing device, 2 studies [46]
[47] used the IOP plication system, and 1 study [62] used the Endomina plication system to perform EGR. Mean age and BMI of the intervention
arm ranged from 38 to 47 years and from 34.8 to 36.2 kg/m2, respectively (Supplementary Table 9). The intervention arm of all studies underwent concomitant LM (moderate intensity
for all studies except for Sullivan et al [47], which underwent concomitant low-intensity LM). The control arm of Sullivan et al
[47] underwent a sham procedure with concomitant low-intensity LM, whereas in the remaining
studies moderate-intensity LM alone was used (Supplementary Table 9).
Benefits
Three RCTs informed the outcome of percentage of TWL at 12 months [45]
[46]
[47]. Three hundred forty subjects were in the EGR plus LM group and 245 in the LM group.
The MD, representing the difference between the pooled percentage of TWL in the EGR
arm minus the control arm at 12 months, was 8.0 % TWL (95 % CI, 3.4–12.6) in favor
of the intervention (Supplementary Fig. 15, available online). This represented a 4.4 times greater weight loss in the EGR arm
compared with the control arm (pooled weight loss of 10.5 % TWL in the EGR arm vs
2.4 % TWL in the control arm).
A separate meta-analysis including only observational studies was conducted. Twenty-one
studies with 5250 patients reported percentage of TWL at 12 months after EGR and were
included [57]
[74]
[100]
[101]
[102]
[103]
[104]
[105]
[106]
[107]
[108]
[109]
[110]
[111]
[112]
[113]
[114]
[115]
[116]
[133]
[134]. Of these, 16 studies (4880 patients) used the Overstitch suturing device, 4 studies
(319 patients) used the IOP plication system, and 1 study (51 patients) used the Endomina
plication system to perform EGR. Mean age ranged from 34 to 56 years and BMI from
32.5 to 49.9 kg/m2. At 12 months, the pooled average weight loss was 17.3 % TWL (95 % CI, 16.2–18.4)
(Supplementary Fig. 16A, available online). A subgroup analysis based on the device demonstrated the efficacy
of EGR performed using the Overstitch endoscopic suturing device, IOP, and Endomina
plication system to be 18.2 % TWL, 16.5 % TWL and 7.0 % TWL, respectively, at 12 months
(Supplementary Fig. 16B).
Harms
Four RCTs informed the outcome of SAEs [45]
[46]
[47]
[62]. SAEs were defined by the investigators and reported in the original studies. The
pooled estimate for SAEs showed a relative risk of 5.6 (95 % CI, 1.1–30.1) when comparing
the EGR group (14/435) with the control group (1/253) (Supplementary Fig. 17, available online). Selected examples of SAEs from the ESG study included abdominal
abscess treated with endoscopy (1/131), upper GI bleeding managed conservatively (1/131),
and malnutrition treated with endoscopic reversal of the ESG (1/131). Selected examples
of SAEs from the largest plication ESG study included extraluminal bleeding treated
with laparoscopy (1/221), hepatic abscess treated with percutaneous drainage (1/221),
and abdominal pain, nausea, or vomiting requiring prolonged hospitalization (9/221)
(Supplementary Table 13, available online).
Certainty of evidence assessment
The overall certainty in the evidence of effects for EGR was moderate (Supplementary Tables 11 and 14 and Supplementary Fig. 18, available online). For benefits, we found indirectness for weight loss, making us
rate the certainty in evidence down to moderate. Specifically, whereas Abu Dayyeh
et al [45] used the current technique with placing stitches in the gastric body to reduce its
volume, Miller et al [46] and Sullivan et al [47] used the former technique, which focused on placing plications in the fundus. This
difference in techniques likely explained inconsistency and imprecision of the MD
in weight loss. Additionally, the control group in Sullivan et al [47] underwent a sham procedure with concomitant low-intensity LM, which has been shown
to be associated with a smaller MD in weight loss compared with a non-sham control
group. For harms, there was moderate certainty in evidence given a small number of
SAEs with a wide CI.
Discussion
This analysis included several types of devices for gastric remodeling including the
Overstitch suturing device, IOP plication device, and Endomina plication device. Although
these devices create tissue plications differently, the result is similar. All procedures
reduce the width and length of the stomach and are believed to delay gastric emptying
[74]
[123]
[124]. Currently, the Overstitch has a CE mark and FDA De Novo marketing authorization
for the treatment of obesity, whereas the IOP and Endomina have a CE mark and FDA
510(k) clearance for tissue approximation of the GI tract.
For EGR, the MD in weight loss, representing the difference between the pooled percentage
of TWL in the EGR arm minus the control arm, at 12 months was 8.0 % TWL (95 % CI,
3.4–12.6) in favor of the intervention. The certainty of this evidence was rated moderate.
Variability was seen across the 3 RCTs on EGR likely because of several factors. First,
the trial with the lowest weight loss in the intervention arm was a sham-controlled
study (4.95 % ± 7.04 % TWL). Within that trial, a lead-in group of 34 subjects who
were unblinded to their treatment achieved 40 % more weight loss than the treatment
patients who were blinded to study assignment [47]. Additionally, the same technique was used in a different trial included in the
analysis. Treatment patients achieved significantly more weight loss in this open-label
RCT (13.0 %; 95 % CI, 10.3–15.8) [46], supporting the hypothesis that the sham study design artificially reduces weight
loss in the treatment arm of an EBMT study. Including the randomized sham-controlled
study therefore may have artificially lowered the weight loss compared with what can
be expected in clinical practice but is the most conservative analysis.
Four RCTs with at least 6 months of data were included in the safety analysis with
a low SAE rate of 3.2 %. Additionally, some of these SAEs were because of accommodative
symptoms of nausea and vomiting causing dehydration and abdominal pain, which were
short-lived and resolved without sequelae.
Similar to IGBs, the panel agreed that EGR currently reduces equity solely because
it is not covered by the national health system or insurance in most countries. Therefore,
in most countries only patients who can pay out of pocket have access to this therapy.
Equity would substantially increase by expanding options and accessibility to a wider
range of patients with obesity, including the under-represented minority patients
with obesity, and if this procedure was covered universally by national health systems
and insurance companies. The panel also agreed that acceptability of endoscopic suturing/plication
remodeling of the stomach is high among patients seeking obesity treatment.
In adults undergoing EGR, the ASGE–ESGE suggests the use of antiemetics periprocedurally.
(Conditional recommendation, very low certainty)
Further details regarding the rationale for this recommendation including the results
of systematic reviews, expert survey, and evidence profile are presented in Appendix
5 (available online).
In adults undergoing EGR, the ASGE–ESGE suggests the use of pain medications periprocedurally.
(Conditional recommendation, very low certainty)
Further details regarding the rationale for this recommendation including the results
of systematic reviews, expert survey, and evidence profile are presented in Appendix
6 (available online).
In adults undergoing EGR, the ASGE–ESGE suggests the use of short-term antibiotics
periprocedurally.
(Conditional recommendation, very low certainty)
Further details regarding the rationale for this recommendation including the results
of systematic reviews, expert survey and evidence profile are presented in Appendix
7 (available online).
In adults undergoing EGR, the ASGE–ESGE suggests the use of short-term PPIs after
the procedure over no PPIs.
(Conditional recommendation, very low certainty)
Further details regarding the rationale for this recommendation including the results
of systematic reviews, expert survey and evidence profile are presented in Appendix
8 (available online).
In adults with obesity, the ASGE–ESGE suggests treatment with AT plus LM over LM alone
depending on device availability.
(Conditional recommendation, low certainty)
Further details regarding the rationale for this recommendation including the results
of systematic reviews, meta-analyses, and evidence profile are presented in Appendix
9 (available online).
In adults with obesity, the ASGE–ESGE recommends treatment with TPS only in the context
of a clinical trial.
(No recommendation, knowledge gap)
Summary of the evidence
One RCT assessing the safety and efficacy of TPS was used to inform this PICO [49]. The study included subjects with class I obesity with at least 1 comorbidity and
class II obesity with or without a comorbidity. Mean age and BMI of the intervention
arm were 43 years and 36.8 kg/m2, respectively. The intervention arm underwent concomitant moderate-intensity LM,
whereas the control arm underwent a sham procedure with concomitant moderate-intensity
LM (Supplementary Table 9).
Benefits
One RCT informed the outcome of percentage of TWL at 12 months [49]. One hundred eighty-one subjects were in the TPS plus LM group and 89 in the sham
plus LM group (Supplementary Table 9). The MD, representing the difference between the mean percentage of TWL in the TPS
arm minus the control arm at 12 months, was 6.7 % TWL (95 % CI, 4.5–8.9) in favor
of the intervention (Supplemental Fig. 19, available online).
Harms
One RCT informed the outcome SAEs [49]. SAEs were defined by the investigators and reported in the original study. The
SAEs showed an absolute risk of 18 additional SAEs per 1000 subjects (95 % CI, 3–380)
in the TPS group (6/213) compared with the control group (0/89) (Supplementary Fig. 20, available online). These SAEs included esophageal rupture requiring a surgical repair
(1/213), upper abdominal pain/device impaction (1/213), vomiting/device impaction
(1/213), gastric ulcer/device impaction (1/213), device intolerance (1/213), and device
impaction (1/213) (Supplementary Table 15, available online).
Certainty of evidence assessment
The overall certainty in the evidence of effects for TPS was low (Supplementary Tables 11 and 16, available online). Risk of bias was judged as not serious (Supplementary Fig. 21, available online). The only limitation of the efficacy evidence was imprecision
because of a small number of patients included in the study. For harms, there was
a low certainty in the evidence given a small number of SAEs with a wide CI that crossed
the line of no difference.
Discussion
The TPS is a gastric device with FDA approval in the United States; however, it has
not yet been commercialized. Unlike the IGB, it is not a space-occupying device. The
mechanism of action is related to the device causing intermittent gastric outlet obstruction
with the larger portion of the device, bobbing between the antrum and pylorus with
gastric contractions. Because the larger portion of the device is filled with silicone,
it does not have a risk of deflation and has FDA approval for 12 months of dwell time.
However, only 1 RCT was available for analysis of the current generation of the TPS
[49]. One previous pilot study was performed evaluating an earlier design of the device,
but that device was associated with a high rate of ulceration that occurred in 50 %
of patients [127] and necessitated the design change to its current form. The U.S. multicenter randomized
sham-controlled trial demonstrated significant weight loss over sham and a low SAE
rate of 2.8 %, but there were only 213 patients who received the device either in
the active arm or an open-label extension arm and 89 control patients. Moreover, because
the device has not been commercialized, only a few members of the panel had any experience
with the device, and this experience was limited to the study setting. Because of
the insufficient real-world experience with the device, the panel recommended using
this device for treating obesity only in the context of a clinical trial.
In adults with obesity and T2DM, the ASGE–ESGE suggests treatment with the DJBL plus
LM over LM alone.
(Conditional recommendation, moderate certainty)
Implementation considerations
Summary of the evidence
Three RCTs assessing the safety and efficacy of the DJBL were used to inform this
PICO [50]
[51]
[63]. Of these, 3 studies were used to assess safety [50]
[51]
[63], and 2 studies were used to assess efficacy [50]
[51]. In Koehestanie et al [63], the DJBL was implanted for 6 months. Therefore, the efficacy, which is the difference
in HbA1c reduction and percentage of TWL between the 2 arms at 12 months, was not
able to be assessed. Otherwise, both Thompson et al [50] and Ruban et al [51] had the DJBL implanted for 12 months. All studies included subjects with obesity
and concomitant T2DM. Mean age, BMI, and HbA1c of the intervention arm ranged from
49.5 to 53 years, 34.6 to 38.4 kg/m2, and 8.3 % to 8.9 %, respectively. In Thompson et al [50], the intervention arm underwent DJBL implantation and concomitant low-intensity
LM, whereas the control arm underwent low-intensity LM alone (Supplementary Table 9).
Benefits
Two RCTs informed the outcomes of HbA1c reduction and percentage of TWL at 12 months
[91]
[93]. Two hundred ninety-eight subjects were in the DJBL plus LM group and 192 in the
LM group. The MD, representing the difference between the pooled HbA1c reduction in
the DJBL arm minus the control arm at 12 months, was .73 % (95 % CI, .39–1.06) in
favor of the intervention (Supplementary Fig. 4, available online). The MD, representing the difference between the pooled percentage
of TWL in the DJBL arm minus the control arm at 12 months, was 5.4 % TWL (95 % CI,
4.1–6.7) in favor of the intervention (Supplementary Fig. 22).
A separate meta-analysis including the active arm of the RCTs and observational studies
of DJBL studies of the same patient population (obesity with concomitant T2DM) was
previously conducted [128]. Fourteen studies with 412 DJBL patients were included with a median implantation
duration of 33 weeks (range, 12–52). Mean age ranged from 36 to 54 years, BMI from
30.0 to 48.9 kg/m2, and HbA1c from 6.7 % to 9.2 %. At the time of DJBL explantation, the pooled HbA1c
reduction and weight loss were 1.3 % (95 % CI, 1.0–1.6) and 18.9 % TWL (95 % CI, 7.2–30.6),
respectively.
Harms
Three RCTs informed the outcome of SAEs [50]
[51]
[63], which were defined as events that resulted in early explant. In Ruban et al [51], the rate of early explant was not reported. Therefore, the worldwide registry was
reviewed, and the SAEs were categorized based on the AGREE classification and need
for early explantation. The pooled estimate for SAEs showed an absolute risk of 24
additional SAEs per 1000 subjects (95 % CI, 8–59) in the DJBL group (26/331) compared
with the control group (0/232) (Supplementary Fig. 23, available online). Selected examples of SAEs from the U.S. pivotal study (ENDO trial)
included intolerance (8/212), hemorrhage (6/212), hepatic abscess (5/212), DJBL obstruction
(3/212), pancreatitis (2/212), intestinal perforation (1/212), and ulceration (1/212)
(Supplementary Table 17, available online).
Certainty of evidence assessment
The overall certainty in the evidence of effects for DJBL implantation was moderate
(Supplementary Tables 11 and 18 and Supplementary Fig. 24, available online). For benefits, becasue the lower 95 % confidence limit for HbA1c
reduction crossed the minimal clinically important difference of .5 %, the evidence
was rated down for imprecision. The certainty of evidence for percentage of TWL, otherwise,
was rated as high. For harms, there was moderate certainty in the evidence given a
small number of SAEs with a wide CI.
Discussion
As noted in the Introduction, the small bowel plays a role in glucose homeostasis,
and treatments targeting the small bowel likely have effects that are independent
of weight loss. In an effort to mimic the effects of Roux-en-Y gastric bypass where
the duodenum and part of the jejunum are bypassed, more than 1 device has been developed
to bypass the jejunum with or without bypassing other portions of the GI tract. Only
1 of these devices, the DJBL, has been studied in RCTs and was previously approved
for use in Europe with a CE mark that was obtained in 2010. The CE mark was lost in
2017 because of administrative issues and not related to a concern about safety or
efficacy, and efforts are underway to regain approval in Europe. A previous U.S. multicenter
randomized sham-controlled trial was stopped early by the company because of concerns
of hepatic abscesses despite meeting the primary endpoints, but a new multicenter
RCT for FDA approval is ongoing as of the time of writing of this guideline. The DJBL
is also being studied for approval in India.
The magnitude of HbA1c improvement at 12 months in patients with obesity and concomitant
T2DM was evaluated in 2 RCTs with an additional improvement of .73 % (95 % CI, .39–1.06)
above the control. A previous meta-analysis that included a combination of 14 observational
and RCTs with data on glycemic control between 12 and 48 weeks of implantation found
an absolute improvement in HbA1c of 1.3 % (95 % CI, 1.0–1.6) compared with baseline
[128]. In a subgroup analysis of the RCTs with implantation between 12 and 48 weeks, the
additional improvement in HbA1c in the interventional arm was .90 % (95 % CI, .5–1.3)
above the control arm, consistent with the present analysis despite the shorter duration
of device implantation. Although small-bowel therapies are categorized separately
from gastric devices because of their weight loss–independent effects, the DJBL also
has an effect on weight loss. The present analysis demonstrated a difference of 5.4 %
TWL (95 % CI, 4.1–6.7) in the device arm over the control arm.
The rate of SAEs evaluated across 3 RCTs with at least 6 months of device implantation
time was 8.5 %, with a wide CI. The panel noted that the original U.S. multicenter
RCT was stopped early by the company because of a higher than anticipated rate of
hepatic abscesses. An analysis performed by the sponsor found that the high doses
of PPIs used for bleeding prophylaxis in the United States, but not in other countries,
contributed to a biofilm on the device with a high bacterial load. The U.S. multicenter
RCT ongoing at the time of writing of this guideline has several infection mitigation
strategies to reduce hepatic abscesses. Furthermore, given the risks of suboptimal
T2DM management and that only about half of patients with T2DM are able to achieve
glycemic control on medications [129], the panel believed the benefits of the DJBL outweighed the risks.
The panel found no negative effects on equity at the present time solely because the
device is not commercially available at this time. However, if it were commercially
available and not covered by national health systems or insurance companies, it would
decrease equity because of lack of affordability by many patients. Physicians with
experience using the device reported patient acceptability of the device was high
both because of the lowering of the HbA1c during implantation and the durability of
HbA1c change up to 6 months after device removal [128].
In adults with T2DM, the ASGE–ESGE recommends treatment with DMR only in the context
of a clinical trial.
(No recommendation, knowledge gap)
Summary of the evidence
One RCT assessing the safety and efficacy of DMR was used to inform this PICO [64]. The study included subjects with T2DM and BMIs between 24 and 40 kg/m2. Mean age, BMI, and HbA1c of the intervention arm were 58 years, 31.5 kg/m2, and 8.2 %, respectively. The intervention arm underwent concomitant low-intensity
LM, whereas the control arm underwent a sham procedure with concomitant low-intensity
LM (Supplementary Table 9).
Benefits
One RCT informed the outcome of HbA1c reduction at 6 months [64]. Fifty-six subjects were in the DMR plus LM group and 52 in the sham plus LM group. The
MD, representing the difference between the mean HbA1c reduction in the DMR arm minus
the control arm at 6 months, was .3 % (95 % CI, –1.1 to 1.7) in favor of the intervention
(Supplemental Fig. 25, available online).
Harms
One RCT informed the outcome of SAEs [64], which were defined by the investigators and reported in the original study. The
SAEs showed an absolute risk of 15 additional events per 1000 subjects (95 % CI, 3–375)
in the DMR group (2/56) compared with the control group (0/52) (Supplementary Fig. 26, available online). These SAEs included precautionary hospitalization for hematochezia
later found to be because of external hemorrhoids (1/56) and jejunal perforation requiring
surgical repair (1/56) (Supplementary Table 19, available online).
Certainty of evidence assessment
The overall certainty in the evidence of effects for DMR was low (Supplementary Tables 11 and 20, available online). Risk of bias was judged as not serious (Supplementary Fig. 27, available online). The only limitation of the efficacy evidence was imprecision
because of a small number of patients and the lower 95 % confidence limit for HbA1c
reduction crossing the minimal clinically important difference of .5 %. For harms,
there was low certainty given inconsistency because the data were derived from 1 RCT
only and imprecision because of a small number of SAEs with a wide CI.
Discussion
DMR is one of several potential therapies that directly treat the abnormally hypertrophied
small-bowel mucosa that is hypothesized to drive the enteral contribution to poor
glycemic control. The Revita DMR is the only DMR therapy that has undergone an RCT
at this time. A few issues were found with the RCT. The trial was small, with 108
patients randomized to either the active or control arm, and was performed at sites
in Europe and Brazil, which were found to be too heterogenous to be combined into
1 analysis and were stratified by region. Moreover, glycemic control was only reported
out to 24 weeks. In a meta-analysis of single-arm studies, the absolute change in
HbA1c from baseline was 1.72 % (95 % CI, .25–3.19) at 3 months and .94 % (95 % CI,
.68–1.21) at 6 months, with a small change in weight that was not sufficient to explain
the improvement in HbA1c [130]. One single-arm study reported a change in HbA1c of −10 ± 2 mmol/mol at 12 months
in 36 patients [131]. Finally, another small single-arm study performed in biopsy sample–proven nonalcoholic
steatohepatitis patients [132] (11 patients, 82 % of patients with T2DM) found neither significant reduction of
HbA1c nor weight loss reduction.
However, because of the limited number of patients in the RCT, patient heterogeneity
between regions, and only a 24-week study duration, the panel believed the data were
insufficient to make a recommendation for or against DMR in a clinical setting and
that the device should be used in a trial setting only. At the time of the writing
of this guideline, a U.S. and European multicenter RCT evaluating the effect of DMR
on glycemic control is ongoing. This study may provide the additional data needed
to determine whether recommendations should be made for or against this therapy for
the treatment of T2DM.
