Risk Determinants of Type 2 Diabetes Mellitus with Severe Obesity and Prediction Model for Diabetes Remission after Bariatric Metabolic Surgery
Article information
Abstract
Background
Bariatric metabolic surgery (BMS) has been established as an effective intervention for obesity and type 2 diabetes mellitus (T2DM). However, systematic research addressing the onset of diabetes and post-surgical remission in severely obese populations remains scarce. This study aims to identify risk factors for T2DM in populations with severe obesity undergoing BMS and develop and validate a prediction model for the primary outcome of diabetes remission (DR) 1 year after BMS. This research provides a precise tool for managing T2DM in populations with severe obesity.
Methods
This research utilizes the China Obesity and Metabolic Surgery Database, retrospectively analyzing 3,670 severely obese populations who underwent BMS between January 2014 and January 2024. Differential analysis identified risk factors for T2DM onset, while univariate and multivariate regression analyses identified independent risk factors for DR post-surgery. A prediction model for DR was developed and internally validated.
Results
Factors associated with T2DM onset in severely obese populations included family history of diabetes, hypertension, hyperlipidemia, glycosylated hemoglobin (HbA1c) levels, and fasting plasma glucose. Independent factors influencing DR postsurgery included diabetes duration, surgical method, HbA1c, and insulin requirement. Subsequent model validation confirmed stable performance metrics (area under the curve values training, 0.71; validation, 0.72).
Conclusion
This study identifies risk factors for T2DM onset and a prediction model for DR following BMS in the Chinese severely obese population. It provides a more precise risk assessment tool for patients with severe obesity and T2DM, and lays the groundwork for future multicenter studies and international collaborations.
Highlights
• In severe obesity, T2DM risks include family history, hypertension, and hyperlipidemia.
• Diabetes remission is higher after Roux-en-Y gastric bypass than sleeve gastrectomy.
• A new nomogram predicts diabetes remission after surgery with acceptable accuracy.
• Longer diabetes duration and insulin use lower postoperative remission rates.
INTRODUCTION
Obesity and type 2 diabetes mellitus (T2DM) have emerged as significant global public health concerns, affecting approximately 9.3% of the adult population worldwide [1]. Excessive fat accumulation leads to insulin resistance and β-cell dysfunction, while T2DM exacerbates appetite and energy metabolism imbalances, further worsening obesity [2]. Notably, over 53.1% of T2DM cases are directly associated with an increase in body mass index (BMI), particularly among individuals with severe obesity (BMI ≥40 kg/m² according to Centers for Disease Control and Prevention) [3], where the prevalence of T2DM is 43%, in contrast, the prevalence among individuals with normal weight is approximately 8% [4,5]. Compared to class I or II obesity, individuals with severe obesity exhibit a heightened risk of metabolic diseases that further escalates with increasing BMI [6]. This association can be attributed to the fact that a BMI of 40 kg/m² or higher typically signifies more severe metabolic dysfunction, leading to a range of metabolic abnormalities, such as hyperglycemia, hyperinsulinemia, and dyslipidemia, as well as multi-organ involvement, including cardiovascular diseases. Consequently, this necessitates heightened attention and proactive intervention [7]. Recently, the rising prevalence of obesity and T2DM has become a growing concern, especially in rapidly developing countries like China. In China, 4% of adults are affected by severe obesity, and the prevalence of T2DM has exceeded 140 million, accounting for 23% of the global T2DM population [4,8]. As these health issues intensify, the limitations of traditional treatment methods are becoming increasingly apparent, making bariatric metabolic surgery (BMS) an increasingly recognized treatment option for severe obesity with T2DM.
BMS has been demonstrated to possess long-term and sustained efficacy in the treatment of severe obesity combined with T2DM. In 2009, the American Diabetes Association recognized BMS as a recommended treatment for populations with obesity and T2DM [9]. Research in the United States has shown that sleeve gastrectomy (SG) and Roux-en-Y gastric bypass (RYGB) result in weight loss rates of 22.8% and 24.1%, and T2DM remission rates of 55.9% and 59.2%, respectively, 1-year post-surgery [10]. Notably, for obese T2DM populations with a BMI ≥40 kg/m² who have not achieved satisfactory results with non-surgical treatments, BMS should be prioritized as the treatment of choice [11].
The efficacy of BMS in controlling severe obesity combined with T2DM can be evaluated using prediction models based on diabetes remission (DR). Nomograms, as integrated risk prediction tools, are widely used not only for cancer survival but also in assessing risks for non-cancer diseases. However, despite the existence of several models predicting the effects of BMS on DR, most studies focus on Western populations. Given that Asian populations tend to have more visceral obesity and higher T2DM prevalence, existing models have limited applicability in these populations [12,13]. Moreover, continuous studies on preoperative conditions and postoperative remission effects are scarce, and long-term follow-up research on the same group of patients is lacking. Most studies focus on the general obese population, with few specific studies targeting the severely obese group, which often faces more complex health issues [14].
To address these issues, this study utilizes a 10-year clinical dataset from the China Obesity and Metabolic Surgery Database (COMSD), which includes 51 variables related to T2DM. These variables encompass multiple dimensions such as lifestyle, clinical symptoms, medical history, comorbidities, blood test results, and family history. By analyzing the risk factors for T2DM in the severely obese population and evaluating the independent impact of BMS on DR in these populations, we provide a more comprehensive accurate risk assessment and prediction model for this group, offering important clinical insights into the onset and remission of T2DM.
METHODS
Patient selection
COMSD encompassing 128 metabolic surgery centers across 64 cities in China [15]. This study focused on severely obese individuals who underwent BMS between January 2014 and January 2024, with follow-up data extending until January 2025. Inclusion criteria were: (1) aged 18 to 65 years; (2) BMI ≥40.0 kg/m²; (3) underwent only SG or RYGB; and (4) had complete preoperative and 1-year postoperative follow-up records. Exclusion criteria included: (1) prior BMS; (2) revision surgery within 1 year postoperatively; (3) severe organ dysfunction; (4) active malignancy; and (5) severe postoperative complications or infections. The decision regarding the type of BMS was made collaboratively by the surgeon and the patient. Initially, 15,127 individuals were screened, and after the selection process, 3,670 met the inclusion criteria. The study flowchart is presented in Fig. 1. This study was approved by the Ethics Committee of the First Affiliated Hospital of Jinan University (KY-2021-102). All participants provided written informed consent, in accordance with the Declaration of Helsinki.
Workflow of the study design. T2DM, type 2 diabetes mellitus; COMSD, China Obesity and Metabolic Surgery Database; BMI, body mass index; SG, sleeve gastrectomy; RYGB, Roux-en-Y gastric bypass; ROC, receiver operating characteristic; DCA, decision curve analysis.
Study variables
Preoperative baseline characteristics were collected, including age, sex, height, weight, BMI, waist circumference (WC), hip circumference, waist-to-hip ratio, smoking history, alcohol consumption, family history of obesity, family history of diabetes, and surgical approach. Patient-reported symptoms and comorbidities were documented, encompassing hair loss, acid reflux, snoring, Helicobacter pylori infection, hypertension, T2DM, cardiovascular disease, fatty liver, sleep apnea syndrome, thyroid disorders, hyperlipidemia, hyperuricemia, gout, arthritis, and acanthosis nigricans. Preoperative biochemical profiles were analyzed, comprising glycosylated hemoglobin (HbA1c), fasting plasma glucose (FPG), fasting plasma insulin (FPI), fasting C-peptide, total cholesterol, triglycerides (TG), high-density lipoprotein, low-density lipoprotein, alanine aminotransferase, aspartate aminotransferase, creatinine, albumin, free triiodothyronine, free thyroxine, and uric acid. T2DM-related parameters were further evaluated, including diabetes duration, hypoglycemic medication use, insulin requirement, insulin resistance assessed by the homeostatic model assessment of insulin resistance (HOMA-IR), quantitative insulin sensitivity check index (QUICKI), lipid accumulation product (LAP), and McAuley index.
Definitions of DR, T2DM-related factors and other comorbidities
DR was defined according to the 2022 consensus statement as HbA1c <6.5% (48 mmol/mol) measured at least 3 months after cessation of glucose-lowering pharmacotherapy [16]. Diagnostic Criteria for T2DM (based on any one of the following): FPG ≥7.0 mmol/L (126 mg/dL), 2-hour plasma glucose during 75 g oral glucose tolerance test ≥11.1 mmol/L (200 mg/dL), HbA1c ≥6.5% (using National Glycohemoglobin Standardization Program [NGSP]-certified methods), FPG ≥11.1 mmol/L (200 mg/dL)+classic hyperglycemic symptoms [17]. According to clinical guidelines, a diagnosis of hypertension requires systolic blood pressure measurements ≥140 mm Hg and/or diastolic readings ≥90 mm Hg [18]. Hyperuricemia is defined as a serum uric acid concentration exceeding 7.0 mg/dL (420 μmol/L), gout diagnosis requires: recurrent arthritis attacks, tophi (ear/joint deposits), imaging showing joint damage/urate crystals [19]. Diagnosis of other comorbidities was based on guidelines or previous studies [20-24]. HOMA-IR was used to assess the degree of insulin resistance, with the formula: HOMA-IR=FPI×FPG/22.5, where FPI is in μU/mL and FPG in mmol/L. QUICKI is another method for evaluating insulin sensitivity, calculated as: QUICKI=1/[log(FPI)+log(FPG)], with FPI in mU/L and FPG in mg/dL. The LAP index reflects the accumulation of lipids in the body, calculated as: LAP (male)=(WC–65)×TG (mmol/L); LAP (female)=(WC–58)×TG (mmol/L). The McAuley index is also used to assess insulin sensitivity, calculated as: McAuley index=exp [2.63–0.28×ln(FPI)–0.31×ln(TG)], where FPI is in mU/L and TG in mmol/L [25].
Statistical analysis
Statistical analysis was conducted using SPSS software version 26.0 (IBM Co., Armonk, NY, USA) and R software version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria). Continuous variables that followed a non-normal distribution were analyzed using non-parametric methods (rank sum tests and Kruskal–Wallis tests), with results reported as median (interquartile range). Categorical data were assessed using chi-square tests, with results presented as number (%). Adjusted odds ratios (OR) and 95% confidence intervals (CI) were calculated. Multivariate logistic regression was employed to adjust for variables with P<0.05 in univariate logistic regression analyses. Data were randomly allocated to experimental and validation cohorts in a 7:3 ratio, and variables were compared. In the nomogram, variable weights are directly derived from the coefficients of multivariable logistic regression. The contribution of each predictor to DR is quantified by its adjusted OR. The regression coefficients (β) are transformed to a scale ranging from 0 to 100 using the following equation: Pointsi=(βi–βmin)/(βmax–βmin)×100. In this equation, βi represents the regression coefficient of a variable, while βmin and βmax denote the smallest and largest coefficients in the model, respectively. The total score is obtained by summing all points, which is then mapped to the bottom axis of the nomogram to determine the predicted probability of DR. The selected variables were utilized as predictors to develop a DR prediction model in the training cohort and validate it in the test cohort. The discriminative performance of the model was assessed through receiver operating characteristic (ROC) curves and the area under the curve (AUC). The relationship and consistency between observed outcomes and predicted probabilities were evaluated using the Hosmer-Lemeshow goodness-of-fit test and calibration curves, with discrepancies reflecting the model’s reliability. Lastly, decision curve analysis (DCA) was conducted to determine the model’s clinical utility and applicability.
RESULTS
Risk factors for T2DM in severely obese populations
From an initial cohort of 15,127 individuals undergoing BMS, 3,670 individuals were ultimately included in this study. Comprehensive comparisons of baseline characteristics, metabolic laboratory parameters, and obesity-related comorbidities were conducted between severely obese populations stratified by T2DM status (Table 1). The T2DM group (n=1,021, 27.8% of the cohort) exhibited significantly higher median WC (131 cm vs. 130 cm, P=0.006) and waist-to-hip ratio (1.03 vs. 1.03, P=0.009) compared to non-T2DM individuals, along with a greater prevalence of smoking (19.4% vs. 16.6%, P=0.043). T2DM patients demonstrated elevated rates of obesity-related comorbidities, including hypertension (64.6% vs. 48.3%, P<0.001), cardiovascular disease (2.4% vs. 1.2%, P=0.008), fatty liver (92.1% vs. 88.8%, P=0.004), sleep apnea syndrome (58.4% vs. 52.5%, P=0.001), hyperlipidemia (69.0% vs. 49.4%, P<0.001), and acanthosis nigricans (29.3% vs. 21.5%, P<0.001). Family history of diabetes was more prevalent in T2DM patients (37.0% vs. 10.0%, P<0.001), who also showed a higher preference for RYGB procedures (24.49% vs. 15.06%, P<0.001). Preoperative biochemical profiles revealed significantly higher median values in T2DM patients for HbA1c (7.10% vs. 5.70%, P<0.001), FPG (7.41 mmol/L vs. 5.49 mmol/L, P<0.001), FPI (27 mIU/L vs. 24 mIU/L, P=0.009), and TG (2.42 mmol/L vs. 1.76 mmol/L, P<0.001). No significant differences were observed in other baseline characteristics, laboratory indices, or comorbidities (P>0.05).
Characteristics of severe obesity population
Risk factors for T2DM
To further elucidate the risk factors for T2DM in the severely obese population, we conducted univariate and multivariate regression analyses on 42 collected variables (Table 2). The results indicated that family history of diabetes was strongly associated with an increased risk of T2DM (adjusted OR=5.17; P<0.001). Additionally, hypertension (adjusted OR=1.89; P<0.001) and hyperlipidemia (adjusted OR=1.54; P<0.001) were identified as independent risk factors for T2DM. Moreover, HbA1c (adjusted OR=1.89; P<0.001), FPG (adjusted OR=1.23; P<0.001), and TG (adjusted OR=1.13; P<0.001) were significant factors associated with the occurrence of T2DM.
Logistic regression analysis of factors associated with presence of diabetes in a severe obesity population
Characteristics of the training and validation cohorts for T2DM patients after BMS
We conducted a further analysis of 1,021 severely obese populations with T2DM. Populations were randomly assigned to a training cohort (714 cases) and a validation cohort (307 cases) in a 7:3 ratio. The median age of all patients was 28 years, with a median BMI of 44.0 kg/m². Male and female patients comprised 40.1% and 59.9% of the cohort, respectively. The majority of patients (75.5%) underwent SG as their primary surgical procedure. The cohort’s median HbA1c and FPG levels were 7.10% and 7.4 mmol/L, respectively. The training and validation cohorts demonstrated comparable baseline and clinical characteristics, with no statistically significant differences (P>0.05) (Supplementary Table 1).
Variable selection and nomogram development
In the training cohort, univariate regression analysis of 51 variables identified six statistically significant predictors (P<0.05). We performed multivariate logistic regression using DR after BMS as the dependent variable, with these six variables serving as independent predictors. At the 1-year follow-up post-BMS, RYGB (adjusted OR=1.88 vs. SG; P<0.001), Hyperlipidemia (adjusted OR=0.66; P=0.016), HbA1c (adjusted OR=0.80; P=0.001), diabetes duration (adjusted OR=0.90; P<0.001) and insulin requirement (adjusted OR=0.59; P=0.006) emerged as independent predictors of DR. Table 3 presents the complete results from both univariate and multivariate analyses. As shown in Supplementary Table 2, significant differences were observed between the DR and non-DR groups at the 1-year post-BMS mark. The overall DR rate was 49.7%, with subgroup rates of 58.8% for RYGB and 46.7% for SG. Additionally, the DR group exhibited significantly different profiles in hyperlipidemia, HbA1c levels, diabetes duration and insulin requirement compared to non-DR patients (all P<0.05). Based on the multivariate results, we developed a clinically applicable nomogram (Fig. 2) to predict the 1-year probability of DR following BMS in severely obese T2DM populations. The nomogram assigns weighted points (scale 0.1 to 0.7) to each predictive factor, where higher total scores indicate a greater likelihood of DR. Clinicians can sum individual variable scores to obtain total points and directly read the corresponding DR probability from the nomogram’s lower axis.
Variables based on logistic regression analysis (training cohort)
Nomogram of predicted probability of diabetes remission after bariatric surgery. RYGB, Roux-en-Y gastric bypass; SG, sleeve gastrectomy; HbA1c, glycosylated hemoglobin.
Evaluation and validation of the nomogram
The model’s discriminative ability was assessed by calculating the AUC and ROC curves. The training cohort exhibited an AUC of 0.71 (95% CI, 0.67 to 0.75) (Fig. 3A), while the validation cohort demonstrated an AUC of 0.72 (95% CI, 0.66 to 0.77) (Fig. 3B), indicating robust estimation accuracy. Calibration curves were utilized to further validate the model’s fit. Hosmer-Lemeshow tests indicated no statistically significant deviations between the model-predicted probabilities and ideal curves in both the training (P=0.649) and validation cohorts (P=0.173) (Fig. 3C and D), demonstrating excellent goodness-of-fit. The close alignment of the calibration curves with the 45° reference line further confirmed the prediction accuracy. DCA illustrated substantial clinical utility of the model at high-risk threshold intervals (Fig. 3E and F), underscoring its clinical relevance. This nomogram integrates strong discrimination, precise calibration, and actionable clinical value for postoperative diabetes management.
Prediction model performance. The discrimination between the training cohort (A) and the validation cohort (B) was shown by the receiver operating characteristic curves. The calibration curves described the agreement between the predicted probabilities and the observed results based on the training (C) and the validation cohorts (D). The clinical benefits of the training cohort (E) and validation cohort (F) were proven using the decision curve analysis curve. AUC, area under the curve; CI, confidence interval.
DISCUSSION
Obesity and T2DM are intricately linked through their pathogenesis and epigenetic factors, prompting an increasing number of studies to focus on the joint diagnosis and treatment of these conditions as metabolic diseases [26,27]. A nationwide survey in China revealed that the prevalence of T2DM significantly escalates with BMI, demonstrating a positive correlation between BMI and T2DM (hazard ratio=1.50; 95% CI, 1.10 to 2.02; P=0.01) [5,8]. Our study indicated that among individuals with severe obesity (BMI ≥40 kg/m2) undergoing BMS, the prevalence of T2DM reached 23%, aligning with previous research trends [28,29]. Furthermore, T2DM patients with severe obesity exhibited significantly higher WC and waist-to-hip ratios compared to their non-T2DM counterparts. Each 10 cm increase in WC or a 0.1 increase in waist-to-hip ratio corresponded to a 60% heightened risk of T2DM [30]. This phenomenon can be attributed to excessive visceral fat accumulation, which leads to an increased release of free fatty acids (FFA) that impair insulin signaling pathways and disrupt liver glucose-lipid metabolism, ultimately resulting in insulin resistance. Additionally, our data suggest that individuals with T2DM and severe obesity are at a greater risk for complications such as hypertension, heart disease, fatty liver, sleep apnea syndrome, hyperlipidemia, and acanthosis nigricans, with these differences being statistically significant. These pathological characteristics are essentially part of a systemic metabolic disorder caused by insulin resistance and fat deposition, including lipotoxicity, glucotoxicity, and chronic inflammation. The manifestation of these conditions involves complex interactions, including adipose tissue dysfunction, endocrine hormone imbalances, and immune system activation [26,31,32].
To further elucidate the risk factors for T2DM in severely obese individuals, we conducted univariate and multivariate regression analyses. These analyses revealed that family history of diabetes, hypertension, hyperlipidemia, HbA1c, FPG, and TG are independent risk factors for T2DM. Previous studies have indicated that T2DM has a genetic predisposition, with individuals possessing family history of diabetes being up to 2.4 times more likely to develop the condition [33,34]. In our study, the OR was 5.17, which may be attributed to further disturbances in glucose and lipid metabolism within the severely obese population. Hypertension is widely acknowledged as a major risk factor for T2DM, and a significant positive correlation exists between obesity and hypertension. Inflammatory adipokines and insulin resistance induced by obesity contribute to systemic vascular inflammation, impairing vascular relaxation and increasing vascular stiffness, ultimately leading to the development of hypertension [35,36]. Additionally, elevated TG and hyperlipidemia are particularly prevalent in severely obese populations with T2DM. In the severely obese state, the accumulation of adipocytes surpasses the storage threshold, resulting in a substantial release of TG into the bloodstream, which triggers hyperlipidemia. Concurrently, FFAs accumulate in muscle and liver tissues, causing lipotoxicity, further exacerbating insulin resistance, and damaging these tissues [37]. HbA1c and FPG are critical indicators reflecting long-term and immediate blood glucose levels, playing an essential role in the diagnosis, evaluation, and monitoring of T2DM treatment [9].
Severely obese individuals often face significant challenges in achieving ideal weight loss through traditional treatments, which has led to the increasing preference for BMS as a therapeutic option. Currently, multicenter studies examining the postoperative DR outcomes of BMS in patients with severe obesity and T2DM are relatively limited, with the majority of existing reports focusing on Western countries. We further explored T2DM patients within severely obese populations and included variables related to diabetes and BMS in both univariate and multivariate regression analyses. The results indicated that HbA1c, diabetes duration, surgical procedure, hyperlipidemia, and insulin requirement were identified as independent predictors of DR. To validate our findings, we developed a corresponding nomogram and evaluated the model’s stability and accuracy using the AUC and calibration curves derived from both training and validation cohorts. Our analysis demonstrated that our prediction model exhibits favorable predictive capability.
BMS, by modifying the anatomy and function of the gastrointestinal tract, activates various metabolic mechanisms that influence glucose metabolism, insulin sensitivity, and hepatic glucose production. It is recognized as one of the most effective strategies for treating obesity and T2DM [38]. Among the various surgical options, RYGB and SG are currently the most commonly employed procedures. While some controlled studies have shown no significant difference in DR rates between RYGB and SG [39,40], our study revealed that the DR rate in the RYGB group was 1.89 times higher than that in the SG group, with remission rates of 58.8% and 46.7%, respectively. The superior efficacy of RYGB over SG in terms of both DR and weight loss has been supported by several studies [38,41,42]. Furthermore, additional research indicates that the likelihood of T2DM recurrence after SG is 5.5 times higher than that after RYGB [41]. This discrepancy may be attributed to the mechanisms by which RYGB addresses severe obesity in conjunction with T2DM. RYGB reduces gastric capacity and alters intestinal nutrient absorption pathways, thereby promoting the rapid secretion of gut hormones such as glucagon-like peptide-1 and gastric inhibitory polypeptide, which enhance insulin sensitivity [38]. Based on these findings, RYGB should be considered the preferred surgical option for severely obese populations with T2DM, as recommended by current guidelines [11]. Additionally, particularly in the context of severe obesity combined with T2DM, the cost-effectiveness model for RYGB demonstrates a higher return on investment [43]. Studies have identified HbA1c, diabetes duration, and preoperative insulin requirement as independent predictors of DR, with these factors being closely related to β-cell function. Specifically, diabetes duration is negatively correlated with DR, as chronic insulin resistance leads to prolonged exposure of pancreatic β-cells to glucotoxicity and lipotoxicity, impairing their functionality [44-46]. Similarly, hyperlipidemia also affects insulin secretion from β-cells, leading to insulin resistance, reduced insulin sensitivity, and exacerbated inflammatory responses, all of which play crucial roles in the presence and remission of diabetes [47]. Therefore, for patients with these risk factors, postoperative management should include enhanced blood glucose monitoring and, when necessary, medication or insulin therapy.
Several prediction models have been developed to assess the risk of DR after BMS. Among these, the most commonly utilized models include type 2 diabetes remission (DiaREM), Advanced-DiaRem (AdDiaREM), and the age, body mass index, C-peptide level, and duration of T2D (ABCD). The DiaREM model was constructed using four preoperative variables selected from a total of 259 clinical variables: age, HbA1c levels, type of antidiabetic medication, and insulin requirement [12]. The AdDiaREM model expands upon the DiaREM by incorporating two additional variables—diabetes duration and antihypertensive medication use—while assigning different weights to each variable to enhance the accuracy of DR predictions post-BMS [13]. However, due to physiological differences between Asian and Western populations, the applicability of these models to Asian populations is somewhat limited. The ABCD model, which is based on age, BMI, C-peptide levels, and diabetes duration, has primarily been applied in research involving populations with lower BMI [48]. Additionally, all of the aforementioned models were developed for single-surgery types. A recently developed predictive model, the age, BMI, insulin use, duration (ABID) scoring system, has been designed to assess the likelihood of T2DM remission following metabolic surgery, specifically in Asian patients with a BMI <32.5 kg/m². This study demonstrates that age, BMI, insulin use, and duration of T2DM emerge as the key predictive factors for remission. However, the model’s applicability is currently confined to individuals with mild obesity [49]. In response to these limitations, the present study incorporated two types of BMS and constructed a DR prediction model specifically tailored for severely obese Asian populations.
This study presents several limitations. Firstly, it only conducted internal validation and did not include external validation, which may result in overfitting of the model and compromise its generalizability across diverse populations. Secondly, as a retrospective study, it lacks data on important factors such as the type of diabetes medication and the severity of diabetes, rendering it vulnerable to selection bias and recall bias. Consequently, future research should explore the following avenues for expansion: first, incorporating additional potential confounders, to enhance the accuracy of the prediction model; second, we aim to conduct prospective studies and engage in international multicenter collaborations to validate the model’s universality and stability. Additionally, future multicenter studies with extended follow-up (>5 years) are warranted to evaluate long-term sustainability and recurrence rates, as well as to include a separate analysis or subgroup model for individuals with lower BMI.
In conclusion, this study analyzed data from 51 independent variables across 128 weight loss centers in China to identify the risk factors for the presence of diabetes in severely obese individuals and to establish a predictive model for DR following BMS in individuals with severe obesity and pre-existing diabetes. After internal validation, the developed nomogram exhibited exceptional accuracy and reliability. This research provides valuable insights for clinicians, and offers a precise prognostic assessment tool that serves as an important reference for future clinical practice.
SUPPLEMENTARY MATERIALS
Supplementary materials related to this article can be found online at https://doi.org/10.4093/dmj.2025.0337.
Characteristic and group information of diabetic patients
Patient characteristics of diabetes remission after surgery
Notes
CONFLICTS OF INTEREST
No potential conflict of interest relevant to this article was reported.
AUTHOR CONTRIBUTIONS
Conception or design: Z.W., Y.L., D.W.
Acquisition, analysis, or interpretation of data: all authors.
Drafting the work or revising: Z.W., Y.L., D.W., C.W.
Final approval of the manuscript: all authors.
FUNDING
This study was funded by the Science and Technology Projects in Guangzhou (funding no. 202201020063) and the flagship specialty construction project-General surgery of The First Aff iliated Hospital of Jinan University (funding no. 711003).
ACKNOWLEDGMENTS
None