Biliopancreatic Limb Length as a Potential Key Factor in Superior Glycemic Outcomes After Roux-en-Y Gastric Bypass in Patients With Type 2 Diabetes: A Meta-Analysis

Compared with medications and lifestyle modifications, metabolic surgery is superior for controlling type 2 diabetes (1,2). Metabolic surgery is recommended for patients with obesity and type 2 diabetes according to clinical guidelines (3). Roux-en-Y gastric bypass (RYGB) is one of the most effective metabolic surgery procedures for glycemic control and is considered the gold standard for type 2 diabetes surgical treatment (4). The main characteristic of RYGB is the creation of bypass limbs of the foregut, consisting of a biliopancreatic limb (BP limb) and a Roux limb and presumed to be an important factor for postoperative glycemic control (5). However, there is a paucity of literature on determining the optimal length of each limb for improved glycemic control. Consequently, the lengths of each limb used by surgeons vary widely, from 10 cm to 250 cm (6). Considering that bariatric surgery has evolved into metabolic surgery, which is primarily intended for type 2 diabetes control, it is imperative to investigate whether there are ideal limb lengths in RYGB for improved glycemic outcomes.

Weight loss after RYGB is a possible mechanism underlying postoperative glycemic control (5,7). In addition, exclusion of the foregut from nutrient transit in itself results in weight loss–independent glucose-lowering effects (8,9). Although the lengths of the BP limb and Roux limb are expected to affect the glucose metabolism, glycemic outcomes of various BP or Roux limb lengths after RYGB have not been well studied. Uncovering the differential effects of BP and Roux limb lengths in RYGB on glycemic control may spur efforts to find the optimal length for each limb and could suggest novel hypotheses to explain the mechanisms underlying glycemic control after RYGB.

In this study, we performed a systematic review and meta-analysis to investigate how the lengths of the BP and Roux limbs in RYGB differentially affect postoperative glycemic outcomes in patients with obesity and type 2 diabetes and to suggest optimal BP and Roux limb lengths to improve postoperative glycemic outcomes.

This study was performed and reported based on the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) 2009 statement (10). The protocol was registered in the International prospective register of systematic reviews (PROSPERO) (reg. no. 288612). This study was approved by the Institutional Review Board of Korea University Anam Hospital, Seoul, South Korea (no. 2022AN0157).

PubMed, Embase, and the Cochrane Library databases were searched from their inception to 1 October 2021 for relevant studies. The search terms used, with adaptation for each database, and the full search strategy for each database are listed in Supplementary Table 1.

The following studies were included: studies of patients with type 2 diabetes who underwent RYGB with or without other study arms, studies with the lengths of the BP and Roux limb specified, and studies with the glycemic outcomes (e.g., glycemic status defined in terms of remission or improvement) after RYGB presented as rates of patients, primary or secondary outcomes, or with data allowing calculation of the rates of patients achieving each study-specific glycemic outcome. The exclusion criteria were as follows: studies with undefined glycemic outcome (e.g., type 2 diabetes remission or improvement); studies on adolescents or pregnant women; studies with patients’ baseline BMI <35 kg/m²; studies on gastric onco-metabolic surgeries, which are intended to improve type 2 diabetes; studies on revision RYGB after initial bariatric surgery; and case reports, reviews, and editorials and studies not written in English. When studies with overlapping subjects were identified, we chose the studies with more comprehensive data, according to consensus.

Two authors (Y.K. and S.L.) independently assessed the eligibility of all studies retrieved from the electronic literature search, based on titles and abstracts. The screened studies were subjected to full-text review and evaluation. Disagreements between authors were resolved by discussion for consensus.

Two independent authors (Y.K. and S.L.) extracted the following data: first author’s name, the year of publication, study location, study design, length of BP and Roux limb (in centimeters), rate and definition of glycemic outcome (e.g., remission or improvement) after RYGB, baseline age, baseline BMI, sex, number of study participants, baseline glycated hemoglobin (HbA_1c), baseline fasting plasma glucose (FPG), percentage excess weight loss (%EWL), and duration of type 2 diabetes. We assessed the risk of bias and study quality using the Newcastle-Ottawa Scale (11).

We performed meta-analyses using two different definitions of glycemic outcomes: remission and improvement. We defined “type 2 diabetes remission” as HbA_1c <6.0% and FPG <100 mg/dL for at least 1 year, without antidiabetes medication (12). We also defined “type 2 diabetes improvement” as a reduction in HbA_1c, FPG, or number or doses of antidiabetes medications, including type 2 diabetes remission. Regardless of the specific terms used in each study, we classified studies into categories of type 2 diabetes remission or improvement based on our definition.

We classified the included studies into four groups according to the length of the two limbs: short BP–short Roux, short BP–long Roux, long BP–short Roux, and long BP–long Roux (Supplementary Fig. 1). We chose a near-median value (100 cm) as the cutoff for dividing the BP and Roux limbs into long and short groups. The near-median value was chosen according to the distribution of the number of studies included in each group equally to maximize the statistical power of the comparison. We performed an incidence meta-analysis with the restricted maximum likelihood method to pool the rates of type 2 diabetes remission or improvement in individual studies by group. For the sensitivity analyses, we also defined the total limb length as the sum of the BP and Roux limb lengths. Type 2 diabetes remission or improvement rates were also calculated in the four groups as follows: short BP–short total limb, short BP–long total limb, long BP–short total limb, and long BP–long total limb. We chose 200 cm as the cutoff for dividing the long and short total limb groups according to the median values of the included studies. Statistical significance was defined as a two-tailed P < 0.05. Cochran Q test and I² statistics were used to assess heterogeneity between the included studies. I² values of 25%, 50%, and 75% were considered to indicate low, moderate, and high heterogeneity, respectively. Funnel plots were used to visualize publication bias, and Egger test was used to measure the asymmetry of the funnel plot, with a level of significance of P < 0.10.

To investigate whether there was heterogeneity or group differences in baseline characteristics among the four groups based on BP and Roux limb length, we performed Cochran Q test for heterogeneity to test group differences and calculated I² statistics. The baseline characteristics were age, BMI, HbA_1c, FPG, duration of diabetes, and insulin usage, and postoperative %EWL was also investigated. We performed meta-regression to investigate the influence of the two limb length statuses in RYGB on the heterogeneity of type 2 diabetes remission or improvement rates between studies. The moderators used in the meta-regression were Roux limb length, BP limb length, and the BP limb length–to–total limb length ratio. We performed meta-regression using a random-effects model and the restricted maximum likelihood method with log transformation for every moderator to achieve a linear correlation between type 2 diabetes remission or improvement and moderators and to improve the normality of their distributions (13). All statistical analyses were performed with STATA 16.1 software (StataCorp, College Station, TX).

Of the 4,094 articles found in the literature search, 2,822 articles remained after exclusion of duplicates. After further exclusion of 2,558 articles following the initial screening, the remaining 264 articles were subjected to full-text reading, and 28 articles (2,14–40) were finally included. The full literature selection process is illustrated in Supplementary Fig. 2. The included studies comprised 4,509 patients (Table 1). Their average age ranged from 34.8 to 52.4 years and average baseline BMI ranged from 37 to 49.1 kg/m². The BP limb length ranged from 30 to 200 cm, and the Roux limb length ranged from 50 to 150 cm. There were 9 studies (21,24,25,30,34–37,40) with reporting of type 2 diabetes remission rates with remission criteria that met our type 2 diabetes remission criteria, and the remaining 19 studies (2,14–20,22,23,26–29,31–33,38,39) met our type 2 diabetes improvement criteria. All but one (28) of the included studies included evaluation of glycemic outcomes 12 months after RYGB.

In the quality assessment, the “selection of the non-exposed cohort” term was not evaluated, since this study was performed in the fashion of incidence meta-analyses, and there were few studies with direct comparison of the effect of differential limb lengths for the two limbs on glycemic outcomes after RYGB (Supplementary Table 2). None of the studies received reduced scores, implying a minimized risk of bias.

The overall type 2 diabetes remission rate 1 year after RYGB in the nine studies was 0.68 (95% CI 0.57–0.78, I² = 74.4%) (Fig. 1A and Supplementary Fig. 3). The type 2 diabetes remission rate was as follows in the four groups, in descending order: long BP–long Roux group, 0.80 (95% CI 0.70–0.90, I² = 0.0%); long BP–short Roux group, 0.76 (95% CI 0.66–0.87, I² = 0.0%); short BP–long Roux group, 0.57 (95% CI 0.36–0.78, I² = 84.5%); and short BP–short Roux group, 0.62 (95% CI 0.43–0.80, I² not assessable). The group difference test did not show statistically significant differences among the four groups (P for heterogeneity = 0.12).

The overall type 2 diabetes improvement rate 1 year after RYGB in the 28 studies was 0.68 (95% CI 0.62–0.74, I² = 89.9%) (Fig. 1B and Supplementary Fig. 3). The type 2 diabetes improvement rates in the four groups, in descending order, were as follows: long BP–short Roux group, 0.82 (95% CI 0.75–0.89, I² = 0.0%); long BP–long Roux group, 0.81 (95% CI 0.73–0.89, I² = 73.8%); short BP–long Roux group, 0.64 (95% CI 0.53–0.75, I² = 91.5%); and short BP–short Roux group, 0.53 (95% CI 0.45–0.61, I² = 44.3%). The long BP–long Roux and long BP–short Roux groups showed 17% and 31% higher type 2 diabetes improvement rates than did the short BP–long Roux and short BP–short Roux groups, respectively. The test of group differences was statistically significant among the four groups (P for heterogeneity <0.01). No publication bias was detected for any outcome in the funnel plots or Egger test (Supplementary Fig. 4).

Heterogeneity and group difference tests for baseline clinical characteristics and postoperative %EWL according to the four groups defined by BP/Roux limb lengths are shown in Table 2. The four groups did not differ significantly in five baseline characteristics (baseline age, BMI, HbA_1c level, FPG level, and type 2 diabetes duration) or postoperative %EWL. The proportion of insulin use at baseline did not differ significantly among the groups. The I² statistics among groups were also <25% for all six factors, indicating a low level of heterogeneity) (Supplementary Figs. 5–11).

Additional meta-analysis with four groups categorized by BP limb and total limb length showed that the type 2 diabetes remission rates of the short BP–short total limb, short BP–long total limb, long BP–short total limb, and long BP–long total limb groups were 0.67 (95% CI 0.58–0.75, I² = 0.0%), 0.30 (95% CI 0.17–0.44, I² not assessable), 0.73 (95% CI 0.54–0.91, I² not assessable), and 0.79 (95% CI 0.72–0.87, I² = 0.0%), respectively (Supplementary Fig. 12). The group difference test showed a statistically significant difference among the four groups (P for heterogeneity <0.01).

The type 2 diabetes improvement rates of short BP–short total limb, short BP–long total limb, long BP–short total limb, and long BP–long total limb groups were 0.62 (95% CI 0.54–0.70, I² = 85.5%), 0.53 (95% CI 0.09–0.97, I² = 95.8%), 0.84 (95% CI 0.75–0.92, I² = 0.0%), and 0.81 (95% CI 0.74–0.87, I² = 62.5%), respectively. The group difference test showed a statistically significant difference among the four groups (P for heterogeneity <0.01).

The type 2 diabetes remission rate 1 year after RYGB correlated positively with the log of BP limb length (coefficient 0.26, 95% CI 0.03–0.49, P = 0.03) (Fig. 2 and Supplementary Table 3). The type 2 diabetes improvement rate 1 year after RYGB also correlated positively with the log of BP limb length (coefficient 0.24, 95% CI 0.03–0.45, P = 0.02). The type 2 diabetes remission rate 1 year after RYGB had a positive correlation with the log of BP limb length–to–total limb length ratio (coefficient 0.52, 95% CI 0.22–0.83, P = 0.001). The type 2 diabetes improvement rate at 1 year after RYGB was also positively correlated with the log of BP limb length–to–total limb length ratio (coefficient 0.40, 95% CI 0.08–0.73, P = 0.03). The Roux limb length was not significantly correlated with the type 2 diabetes remission or improvement rates at 1 year after RYGB.

Our results showed that the length of the BP limb may be involved in the mechanism of action underlying the superior glycemic outcomes after RYGB in patients with an average BMI >35 kg/m². The long BP group (≥100 cm) had a range of 14%–23% higher type 2 diabetes remission rate than the short BP group (<100 cm), while the long Roux group (≥100 cm) had a range of −5% to 4% increased remission rate as compared with the short Roux group (<100 cm). The long BP group (≥100 cm) had a range of 17%–31% higher type 2 diabetes improvement rate than the short BP group (<100 cm), while the long Roux group (≥100 cm) had a range of −1% to 13% increased improvement rate as compared with the short Roux group (<100 cm). There was no significant difference or heterogeneity in baseline characteristics, including diabetes-related variables and the extent of postoperative weight decrease, among the groups categorized based on BP and Roux limb length. Meta-regression analysis also showed that a longer BP limb was associated with better postoperative glycemic outcomes. This finding supported surgical strategies of creating longer BP limbs rather than longer Roux limbs for improving glycemic outcomes after RYGB. In addition, the efficacy of previous clinical trials for the glycemic outcome of RYGB should be interpreted with caution regarding BP limb length, and subsequent clinical trials on RYGB should address the limb-length status.

We adopted incidence meta-analysis methodology to compare glycemic outcomes according to BP and Roux limb length, as the comparative risk between various limb lengths could not be calculated in the included studies. However, several analytical strategies support the superiority of BP limb elongation, as compared with Roux limb elongation, for glycemic control after RYGB. First, comparison of incidence between independent groups could be influenced by the characteristics of the comparator groups. However, we identified no significant difference or heterogeneity in baseline characteristics among the groups, which implies minimization of confounding bias. Second, meta-regression analysis, which suggested a positive correlation between longer BP limb length and better glycemic outcome, appropriately addressed the concern of bias attributable to arbitrary categorization. Third, regardless of the different definitions of glycemic outcome (e.g., remission or improvement), the results consistently indicated that mainly BP limb length, rather than Roux limb length, contributed to achieving better glycemic outcomes, which alleviates the potential influence of glycemic outcome definitions on the study results.

Several theories have been suggested to explain the effects of the BP limb on glycemic control. First, it has been suggested that nutrients reaching the distal small intestine earlier by a longer BP limb induce greater glucagon-like peptide 1 (GLP-1) release and improve the rate of type 2 diabetes remission (5,7,8). However, recent studies comparing short (50 cm) and long (150 cm) BP limbs showed no significant differences in GLP-1 levels between the two groups (40,41). This result suggests that GLP-1 is unlikely to be the main cause of type 2 diabetes remission with longer BP limbs, suggesting that other gut hormones should be investigated in future (7). Second, the change in serum bile acid physiology is another mechanism that explains type 2 diabetes remission after RYGB. RYGB is associated with elevated bile acid levels in both rodents (42) and humans (43). In addition to their role as surfactants, bile acids act as hormones that influence metabolic processes via receptors such as farnesoid X receptor (FXR) and Takeda G-protein–coupled receptor 5 (TGR5) (44). Glucose homeostasis is improved via the intestinal FXR–GLP-1 axis, and the intestinal microbiome is a suspected mediator of glycemic improvement independent of postoperative weight loss (45). Bile acids also increase GLP-1 secretion via TGR5 in colonic L cells, which may be upregulated by the increased delivery of bile acids to the distal ileum by RYGB (46). However, considering that this mechanism does not distinguish the effect of the BP limb and common channel, further studies are required to clarify the role of bile acid in improving glycemic control via a long BP limb.

Altered gut hormone secretion is also a possible mechanism that contributes to glycemic control after RYGB. Enteroendocrine cells in the stomach and small intestine release hormones such as ghrelin, leptin, cholecystokinin, and peptide YY (47). Gut hormones are involved in endocrine signaling by entering the systemic circulation and affecting peripheral targets such as the brain. The gut-brain axis plays a role in maintaining glucose homeostasis, and gut hormones are significant mediators that influence appetite (48). The alteration of gut hormone secretion following RYGB is linked to glycemic improvement (47). Recent studies investigated the differential release of gut hormones based on nutrient sensing in the gastrointestinal tract, which influences insulin sensitivity (5). Future research on how these gut hormones show differential effects on glucose control and appetite depending on BP and Roux lengths will guide the evolution of RYGB toward improved diabetes surgery and help uncover novel diabetes treatment targets.

Although the value of RYGB with a long BP limb was shown by our study, the extent to which a long BP limb increases nutritional risk should be evaluated. Considering that several bariatric surgeries, such as jejunoileal bypass, are currently not used frequently, at least partly due to nutritional risk (4), it is necessary to address nutritional risk when investigating glycemic outcomes in RYGB with a long BP limb. Results of a previous study with comparison of nutritional risk between a group with a 150-cm BP limb and a 75-cm Roux limb and a group with a 75-cm BP limb and a 150-cm Roux limb showed that there was no significant difference in deficiency of folic acid, vitamin B₁₂, iron, and vitamin D, or in anemia, after RYGB (49). Another study also comparing a 75-cm BP limb and a 150-cm Roux limb group with a 150-cm BP limb and a 75-cm Roux limb group showed that long BP limbs do not result in increased iron, vitamin B₁₂, folate, vitamin D, calcium, or albumin deficiencies at 1 year or 2 years after RYGB (50). However, guidelines strongly warn of nutritional risks in bariatric procedures adopting long small intestine bypasses and emphasize the need for postoperative surveillance of nutritional deficiencies (51,52).

This study had some limitations. First, not all included studies were randomized controlled trials, which might limit the level of evidence suggested by our results. Second, because the included studies included glycemic outcomes ∼1 year after RYGB, long-term effects on glycemic outcomes by long BP limb could not be assessed. Considering that the rate of type 2 diabetes relapse after remission reached 47% up to 12 years post-RYGB (53), further studies are needed to determine whether the effect of a long BP limb on glycemic control can be maintained for an extended period of time. Third, additional risk (e.g., surgical complications and impairment of quality of life) of long BP limbs in RYGB could not be assessed due to the paucity of relevant data. For adoption of a long BP limb to improve glycemic outcomes, risk and benefit profiles should be investigated in future studies. Fourth, different definitions of glycemic outcomes after RYGB might have influenced the results, although the two definitions (remission and improvement) adopted in this study did not change our main observations. Fifth, misclassification due to inaccurate limb length measurements in RYGB could have influenced the results. Reliable and reproducible methods for measuring limb length require investigation in future studies.

In conclusion, use of a longer BP limb in RYGB leads to an improved rate of type 2 diabetes remission and improvement by 1 year after RYGB, as compared with a longer Roux limb. Our results imply that bariatric surgeons should consider allocating longer length to the BP limb rather than the Roux limb to improve glycemic outcomes after RYGB in patients with obesity and type 2 diabetes.

Funding. This work was supported by the Korea Medical Device Development Fund grant funded by the Korea government (the Ministry of Science and ICT, the Ministry of Trade, Industry and Energy, the Ministry of Health & Welfare, the Ministry of Food and Drug Safety) (project nos. 9991007295, KMDF_PR_202012D13-02) and a Korea University grant (to S.P.), and the Basic Science Research Program through the National Research Foundation of Korea (grant no. 2020R1I1A1A01070106) (to Y.K.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.K. designed the study, interpreted data, and wrote the manuscript. S.L. collected data, interpreted data, and wrote the manuscript. D.K. conducted statistical analyses. A.A. interpreted data. S.-H.P. interpreted data. C.M.L. critically revised the manuscript. J.-H.K. critically revised the manuscript. S.P. designed the study and critically revised the manuscript. Y.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.