Introduction

In recent decades, the prevalence of both type 2 diabetes mellitus (T2DM) and obesity has increased progressively.1 This common epidemiological trend has also been reflected in the increase in shared risk factors and has accelerated the clinical course of these diseases.2

The key components of contemporary management of chronic metabolic diseases are lifestyle interventions, pharmacotherapy and metabolic-bariatric surgery. In 2009, the American Diabetes Association included metabolic-bariatric surgery for obesity-associated T2DM in its T2DM treatment guidelines.3 Since then, it has been reported that metabolic-bariatric surgery improves glycemic control and achieves higher rates of remission of T2DM than pharmacological treatment. Findings from numerous controlled clinical trials indicate that metabolic-bariatric interventions show significant efficacy in inducing remission of T2DM. The remission rate 2 years after surgery was reported to be 85%, declining to 50% at the 5-year mark, yet surpassing remission rates observed in the pharmacotherapy group. Remarkably, even at the 10-year follow-up, a substantial proportion of treated patients (37.5%) remained in remission.4–6 Furthermore, bariatric surgery has been linked to a decrease in the occurrence of cardiovascular disease, macrovascular and microvascular complications, and mortality rates in individuals diagnosed with T2DM.7

The most common bariatric surgery techniques are sleeve gastrectomy (SG) and Roux-en-Y gastric bypass (RYGB).8 The use of these techniques has changed over time. Data from the International Federation for the Surgery of Obesity and Metabolic Disorders (IFSO) indicate that SG is currently the most frequently conducted procedure, accounting for 58.2% of cases, followed by RYGB in 26.4% of cases.9 However, a recent study conducted in the UK found that RYGB is used more frequently than SG in patients with T2DM (OR 1.55; 95% CI 1.46 to 1.67).10

In Spain, a study conducted between 2000 and 2014 by the Obesity Group of the Spanish Society of Endocrinology and Nutrition found a progressive decrease in the use of RYGB, associated with a parallel increase in SG.11 Data from a previous Spanish study performed between 2001 and 2010 in patients with and without T2DM showed changes in the use of the various techniques, with a decrease in the use of open RYGB and open gastroplasty and an increase in the use of laparoscopic RYGB and laparoscopic gastroplasty since 2006.12 Nevertheless, despite consensus on the clinical indications for surgery,13 further data are required on the use of bariatric surgery in Spain, trends in the clinical characteristics and hospital outcomes of patients undergoing the procedures, and the effect of COVID-19 on their use in clinical practice.

In light of these data, the aims of our study were as follows: (1) to evaluate temporal trends in incidence, clinical characteristics, complications, and hospital outcomes in patients with and without TDM2 undergoing bariatric surgery in Spain between 2016 and 2022; (2) to compare outcomes of hospitalization between patients with and without TDM2 using propensity score matching (PSM); and (3) to identify the variables associated with poorer clinical outcome among patients with and without TDM2 and to determine the influence of TDM2 on this outcome.

MethodsStudy design

We performed a descriptive observational study based on the analysis of a national hospital discharge registry.

Setting and participants

The data source was the Spanish National Hospital Discharge Database (RAE-CMBD, Specialized Care Activity Registry–Minimum Basic Data Set). The RAE-CMBD collects individualized information on all patients admitted to Spanish public hospitals including sex, age, admission and discharge date, the main diagnosis and a maximum of 20 secondary diagnoses, up to 20 procedures (therapeutic or diagnostic), and discharge destination (discharge, transfer to another health center, and death during admission).14 The RAE-CMBD encodes all this information using the International Classification of Diseases, 10th Revision (ICD-10).

The study period ranged from January 1, 2016 to December 31, 2022. The study population comprised patients with an ICD-10 code for obesity, age ≥35 years, and a code corresponding to a surgical intervention (RYGB or SG) at any position in the procedure fields of the RAE-CMBD. Subsequently, patients with and without T2DM were stratified according to the presence of T2DM at any position in the diagnostic field of the RAE-CMBD.

Patients with T1DM (type 1 diabetes mellitus) codes, patients with abdominal neoplasia, and patients with a lack of data for essential variables (age, sex, date of admission and discharge, and discharge destination) were excluded. Likewise, all patients who underwent laparoscopic gastrectomy were excluded. The ICD-10 codes used to select the study population are shown in online supplemental table 1.

Study variables

The main study variables were the annual incidence of SG and RYGB surgeries stratified by T2DM and the outcome of hospitalization measured in terms of in-hospital mortality (IHM), mean length of hospital stay (LOHS), and the need for admission to the intensive care unit (ICU). A variable “severity” was created to cover patients who died in the hospital or were admitted to the ICU.

The study covariates collected for patients undergoing bariatric surgery were age, sex, and date of surgery. Comorbidity was measured using the diseases included in the Charlson Comorbidity Index (CCI), excluding diabetes. These diseases were identified following the methodology described by Sundararajan et al.15 In addition, the CCI was analyzed as a continuous variable, categorized as CCI=0, CCI=1, and CCI≥2.

Other diseases included asthma, depression, obstructive sleep apnea (OSA), dyslipidemia, hypertension, and COVID-19. Complications of bariatric surgery were also recorded, including gastroesophageal reflux disease (GERD), pulmonary embolism, bowel obstruction, bleeding, sepsis, surgical site infection, and pneumonia.

Invasive mechanical ventilation and non-invasive mechanical ventilation were assessed without considering the positioning of the procedures in the RAE-CMBD. ICD-10 codes for comorbidities, complications, and procedures are shown in online supplemental table 1.

Propensity score matching

A PSM analysis was carried out by pairing each patient bearing a T2DM code with another patient who underwent the same surgery but lacked a T2DM code and whose propensity score, derived from multivariate logistic regression, was identical or closely matched. The calculation of the propensity score involved applying matching conditions such as the year of hospitalization, sex, age, and all comorbidities present on admission.16 The selected matching method was a one-to-one approach using calipers with a width ≤0.2 of the SD of the logit of the PS. The quality of the PSM process was assessed by estimating the absolute standardized difference before and after matching.16 Populations are deemed well balanced when the absolute standardized differences are <10% post-PSM. A love plot was created to visually represent how populations become significantly more comparable after PSM.17

Statistical analysis

We calculated the incidence of SG and RYGB per 100,000 inhabitants with T2DM and patients without T2DM for each of the 7 years under analysis. Population data for the years 2016–2022 were sourced from the Spanish National Institute of Statistics.18 To estimate the population with and without T2DM, we used diabetes prevalence data from the National Health Surveys conducted in Spain in 2016 and 2020, as previously described.12 19 20

Incidence rate ratios (IRRs), along with their corresponding 95% CI, were computed using Poisson regression models. These models were used to compare the age-adjusted and sex-adjusted incidence rates of SG and RYGB between individuals with and without T2DM, as well as among patients with T2DM stratified by sex.

The findings from the descriptive analysis were presented as frequencies and percentages in the case of categorical variables, and as either mean and SD or median and IQR in the case of quantitative variables.

The time trend was evaluated using the Cochran-Mantel-Haenszel statistic. Fisher’s exact test was used to compare categorical variables, while the t-test or Mann-Whitney test was used as necessary to compare continuous variables.

A multivariate logistic regression model was used to detect variables linked with severity among patients undergoing RYGB and SG, taking into account the presence of T2DM. These models included sex, age, year of hospital admission, and comorbidities present at admission as covariates. The outcomes of these models are depicted in the form of ORs alongside their respective 95% CIs.

The statistical analysis was performed using Stata V.14 (Stata, College Station, Texas, USA).

Ethical aspects

The RAE-CMBD database is managed by the Spanish Ministry of Health and is accessible free of charge on request.21 Given the administrative and anonymous nature of the registry, and, in accordance with Spanish Law, individual written consent of subjects is not required on admission to the hospital. Similarly, the approval of an ethics committee is not required.

Results

A total of 32,176 bariatric procedures were performed on obese patients aged ≥35 years in Spain between 2016 and 2022. RYGB accounted for 55.52% (n=17,867) and SG for 44.48% (n=14,309). A total of 9287 patients had been diagnosed with T2DM (28.86%). By type of surgery, T2DM was coded in 31.57% of RYGBs (5643/17,876) and 25.53% of SGs (3653/14,309).

Trends in hospitalization and incidence for patients who underwent RYGB and SG

Between 2016 and 2022, RYGB surgeries increased significantly (p<0.05), and those for SG decreased significantly (p<0.05) among men and women with T2DM, as shown in table 1. In men without diabetes, the frequency of hospitalization for SG increased, whereas that for RYGB decreased; in women without diabetes, the frequency of RYGB increased and that of SG decreased (all p<0.05) (table 1).

Table 1

Trends in number of Roux-en-Y gastric bypass and sleeve gastrectomy procedures for patients with and without T2DM according to sex in Spain from year 2016 to 2022

For both types of surgery and both sexes, the number of interventions decreased considerably in 2020 and then recovered in subsequent years.

Analysis of changes in the incidence of bariatric surgery revealed that the frequency of both RYGB and SG in patients with and without T2DM increased between 2016 and 2022 (p<0.001) (figure 1). The incidence in both procedures in 2020 decreased to values lower than in 2016 in patients with and without T2DM.

Figure 1

Changes in Roux-en-Y gastric bypass (RYGB) and sleeve gastrectomy (SG) procedures. Annual incidence rates in patients with and without type 2 diabetes expressed per 100,000 inhabitants (Spain, 2016–2022). *p<0.05 (Poisson regression analysis). T2DM, type 2 diabetes mellitus.

The incidence of bariatric surgery was significantly higher in patients with T2DM than in those without T2DM for all the years analyzed (p<0.001) (figure 1). The Poisson regression–based age-adjusted and sex-adjusted IRR was 4.07 (95% CI 3.95 to 4.20) for RYGB in people with T2DM compared with those without and 3.02 (95% CI 2.92 to 3.14) for SG.

When patients with T2DM were stratified by sex, the incidence of bariatric surgery was higher in women than in men (IRR 1.89, 95% CI 1.79 to 2.00 for RYGB and IRR 1.72; 95% CI 1.61 to 1.84 for SG). These higher incidence figures in women were also observed in patients without diabetes (IRR 2.81, 95% CI 2.69 to 2.93 for RYGB and IRR 2.26; 95% CI 2.17 to 2.36 for SG).

By type of bariatric surgery, the incidence of RYGB was higher than that of SG in patients with T2DM (26.68 vs 17.30 per 100,000 inhabitants with T2DM) and in those without diabetes (6.56 and 5.72 per 100,000 inhabitants without diabetes, respectively).

Clinical characteristics and hospital outcomes for patients who underwent RYGB

As can be seen in table 2, among the patients who underwent RYGB, women accounted for 62.89% of those with T2DM and 75.66% of those without diabetes (p<0.001). Prior to PSM, mean age was significantly higher among patients with T2DM than in those without diabetes (51.87 years vs 48.56 years; p<0.001), and patients with T2DM also had a higher mean CCI (0.49 vs 0.37; p<0.001) and more specific chronic conditions (eg, congestive heart failure, myocardial infarction, peripheral vascular disease, liver disease, kidney disease, dyslipidemia, hypertension, and OSA).

Table 2

Comparison of characteristic, comorbidities, bariatric complications and hospital outcomes among patients with and without T2DM who underwent a Roux-en-Y gastric bypass in Spain from 2016 to 2022, before and after PSM

Compared with patients with T2DM, patients without diabetes more frequently had GERD (13.38% vs 9.94; p<0.001) and sepsis (0.43% vs 0.12; p<0.001) during admission. The median LOHS was equal irrespective of the T2DM status (3 days). The crude IHM was 0.07% for patients with T2DM and 0.1% for patients without (p=0.574). Severity was 9.55% in patients with T2DM and 9.34% in patients without T2DM (p=0.662).

After PSM, sepsis remained more frequent among patients without T2DM, and IHM and severity became similar in both groups (table 2).

The love plot (online supplemental figure 1) shows that PSM resulted in a good fit for the two populations compared.

Clinical characteristics and hospital outcomes for patients who underwent SG

As in patients undergoing RYGB, significant differences were recorded before PSM between patients with T2DM and without T2DM undergoing SG in sex distribution (60.66% vs 71.42%; p<0.001), mean age (52.59 years vs 48.78 years; p<0.001), and CCI (0.57 vs 0.4 p<0.001). The most prevalent comorbidities among patients with and without T2DM were similar to those described in patients who underwent RYGB.

No significant differences in the prevalence of complications after SG were found between the two groups (table 3).

Table 3

Comparison of characteristic, comorbidities, bariatric complications and hospital outcomes among patients with and without T2DM who underwent sleeve gastrectomy in Spain from 2016 to 2022, before and after PSM

LOHS (3 days) and crude IHM (0.08%) were identical irrespective of the presence of T2DM. However, the crude values for admission to ICU and severity were significantly higher in patients with T2DM (both p<0.001).

After adjustment with PSM, all variables with significant differences between patients with and without T2DM ceased to be significant (table 3). The love plot after the PSM is shown in online supplemental figure 2.

Variables associated with severity in patients with and without T2DM who underwent RYGB and SG: multivariable analysis

Female sex was associated with less severe disease only in patients without T2DM who underwent RYGB and SG. Furthermore, disease was significantly less severe during the period 2018–2022 than in 2016 for both surgeries in both sexes irrespective of diabetes status.

In patients with T2DM who underwent both RYGB and SG, multivariable adjustment showed that severity was associated with the presence of postoperative surgical site infection and the need for invasive mechanical ventilation during admission (table 4).

Table 4

Multivariable logistic regression analysis of the factors associated with severity during hospital admission for Roux-en-y gastric bypass and sleeve gastrectomy procedures in Spain, 2016–2022, according to diabetes status

Finally, T2DM was not associated with the severity of patients who underwent RYGB (OR 0.98; 95% CI 0.86 to 1.11), although T2DM was significantly associated with severity in those who received SG (OR 1.23; 95% CI 1.07 to 1.42).

Discussion

This nationwide study based on the T2DM status of more than 32,000 bariatric surgery interventions performed in Spain between 2016 and 2022 revealed several noteworthy findings. First, in patients with T2DM, the frequency of both bariatric surgery procedures has increased (mainly RYGB). Second, in each year of the study, the procedures were more frequent in patients with T2DM than in patients without. Third, more RYGB and SG procedures were performed on women than on men, regardless of diabetes status. Fourth, the presence of T2DM was associated with severity in SG patients.

Several epidemiological studies have indicated that RYGB was more likely to be performed than SG in patients with T2DM,10 22 possibly because RYGB seems to achieve a greater reduction in weight and for longer time in patients with T2DM than SG.7 23 24 A follow-up study of patients with T2DM in the USA comparing different bariatric surgery procedures showed that patients in the RYGB group had 6.2– 8.1% greater total body weight loss than the SG group at 1 and 5 years than those who underwent SG.25 In addition, relapse of T2DM after remission was less frequent after the use of RYGB than after SG. Aminian et al26 found that, in a patient with T2DM who had been receiving diabetes medication for 5 years prior to bariatric surgery, the projected risk of relapse would be 20% after RYGB and 36% after SG. Additionally, McTigue et al25 agreed that relapse of T2DM was less frequent for RYGB than for SG, with a HR of 0.75 (95% CI 0.67 to 0.84).

In our study, patients with T2DM were older and had more comorbid conditions than patients without diabetes. Older age, duration of diabetes, elevated baseline HbA1c values, high body mass index (BMI), and use of antidiabetic medications are predictive factors that predispose patients to a lower probability of remission of T2DM.27

Our database analysis, consistent with a previous Spanish study,12 revealed that bariatric surgery is more prevalent among patients with T2DM than among those without diabetes. Bariatric surgery has been extensively demonstrated as pivotal in addressing both obesity and T2DM, primarily owing to its long-term effectiveness.1

As in most population-based studies, bariatric surgery was more frequent in women, both with and without T2DM.28–31 Compared with men, women with obesity have lower self-esteem, more frequently experience psychological problems (such as depression) and report a greater impact on sex life and physical functioning, with the result that bariatric surgery is now the treatment of choice.30 31

In our study, we observed that disease was less severe in women without T2DM than in men. In their investigation of patients enrolled in the Michigan Bariatric Surgery Collaborative who underwent primary bariatric surgery from 2006 to 2016, Kochkodan et al32 found that male patients exhibited significantly more preoperative risk factors. These factors included a higher BMI and a greater burden of comorbidities. After surgery, male patients experienced more surgical complications and lost less weight, leading the authors to conclude that men tend to seek treatment later than women.

As expected, there was a decline in the number of procedures performed in 2020 due to the impact of the COVID-19 pandemic.9 Beside the lack of hospital beds and limited access to the ICU, these findings could be explained by the greater morbidity and mortality among individuals with T2DM and severe obesity when affected by COVID-19 than in the general population.33 34 Nonetheless, a multinational cohort study illustrated that with appropriate perioperative protocols,35 bariatric surgery could be performed safely, even amidst the COVID-19 pandemic.

Diabetes is considered a risk factor for intraoperative complications in multiple surgeries, and bariatric surgery is no different.36 Results from a large study of the MBSAQIP database indicated that among patients with T2DM, infectious complications were the most frequent cause of death after RYGB and SG.37 In our study, complications due to infection were associated with severity in both RYGB and SG.

As in other studies, sepsis in patients without diabetes was one of the reasons for the greatest severity after bariatric surgery.38 39 Bruschi Kelles et al39 indicated that 42.7% of deaths after bariatric surgery were due to sepsis. The primary cause of sepsis is often anastomotic leakage, that is, leakage of fluids into the abdominal cavity, which subsequently leads to peritonitis and sepsis.40

Adverse respiratory events are reported in approximately 2% of patients after bariatric surgery.41 In our study, pneumonia was associated with greater severity after SG.

As expected, IHM was very low in patients both with and without diabetes.13 42–45 The reasons given for this low mortality rate are the improved use of preoperative antibiotics and prophylaxis for deep vein thrombosis, shorter hospital stay, the correct procedure and technique, and the care received in the ICU.40 42 45

The principal strength of our study lies in its use of the RAE-CMBD, a comprehensive Spanish national population database spanning a period of 7 years. The methodology has been documented elsewhere, thus enhancing the credibility and reproducibility of our findings. It should be noted that the RAE-CMBD includes practically all hospital admissions in Spain (>95%). Despite the aforementioned strengths, our study is constrained by several limitations. One such limitation is inherent in the fact that the RAE-CMBD is an administrative database, which means that it may not encompass all variables present in the medical records. Therefore, we lack information on diabetes such as duration, treatment, and laboratory results (HbA1c). Additionally, we lack information regarding whether participants opted for bariatric-metabolic surgery specifically to manage obesity or T2DM or for other purposes. Furthermore, the RAE-CMBD does not collect the BMI of subjects either before or after surgery, and this would be valuable information for developing care pathways. Second, akin to similar observational and retrospective studies that depend on ICD codes sourced from large databases, our research might be susceptible to issues of low sensitivity and specificity. This susceptibility could be contingent on the accuracy of clinicians in coding hospital procedures and diagnoses, thereby impacting the quality of the data. Nevertheless, previous studies conducted in Spain and elsewhere have demonstrated the validity of diagnosing diabetes using ICD codes in health administrative databases, as compared with clinical records. These studies concluded that such databases can effectively address research inquiries.45–49 Studies conducted in Spain evaluating the validity of the diagnosis of diabetes using the RAE-CMBD indicated a specificity of 97% (ie, almost all patients with a diabetes code actually have the disease), sensitivities of 55% and 63.7% (indicating that the condition is not encoded in some patients who actually have diabetes), and kappa indexes for agreement of 0.6 and 0.7.48 49 Third, patients who had undergone gastric banding were excluded, as this was a less frequent bariatric intervention than SG and RYGB.9 In our opinion, the exclusion of these cases would not result in a significant impact on our results but should be taken into account for future research. Fourth, we employed an age cut-off of 35 years to minimize the risk of diabetes type misclassification. Studies conducted in Spain using administrative data based on ICD-10 have indicated a risk of misclassification of diabetes type, particularly among younger individuals.50 51 However, previous reports in our country have demonstrated that the prevalence of T2DM is below 0.5% in those under 35 years.52 53 Fifth, a well-recognized limitation of administrative discharge databases is the absence of illness severity data.54 The Clavien-Dindo classification could not be used in our investigation because it incorporates information on pharmacological treatments not captured by the RAE-CMBD.55 Due to this limitation and the unavailability of other widely accepted measures of severity, we opted to develop our own definition of “severity.” Admission to the ICU and IHM have been demonstrated to be strongly associated with severity.56 Finally, re-operations were not included as a complication since the ICD-10 version used by the RAE-CMBD does not allow for the identification of patients who undergo surgery more than once during a single hospital admission. In conclusion, between 2016 and 2022, the number of RYGB and SG procedures performed in Spain rose in patients with and without T2DM. More interventions were performed on patients with T2DM than on patients without T2DM. The most common procedure in patients with T2DM is RYGB. IHM was very low after RYGB and SG, and among patients who underwent SG, those with T2DM more frequently experienced a severe postoperative course compared with subjects without T2DM.