Recent studies report the effect of bariatric surgery on glycaemia control and prevention of type-2-diabetes in obese patients. This study is about the pathophysiological mechanisms associated to these changes.
Recent studies report the effect of bariatric surgery on glycaemia control and prevention of type-2-diabetes in obese patients. This study is about the pathophysiological mechanisms associated to these changes.
Circulating levels of receptors of tumor necrosis factor (TNF-RI, TNF-RII), visfatin, high molecular weight (HMW) adiponectin, and C reactive protein (CRP) in 30 morbidly obese women (body mass index, BMI>40 kg/m2) and 60 normal-weight controls (BMI>25 kg/m2) were analyzed. Morbidly obese were studied at three time-points: before surgery (baseline), and 6 and 12 months after.
After surgery, the levels of TNF-RI, TNF-RII, visfatin, and CRP were significantly lower than its baseline levels, whereas HMW adiponectin was higher. Fasting glucose, insulin, and homeostasis model assessment of insulin resistance (HOMA2-IR) levels were markedly lower postoperatively. High density lipoproteins (HDL) moderately increased, and triglyceride levels had sharply decreased. The study of the predictive value of variables indicated that preoperative levels of TNF-RI and visfatin correlated positively with levels of glucose, insulin, glycosylated hemoglobin A1c, and HOMA2-IR postoperatively, whereas adiponectin levels correlated negatively. Baseline CRP levels negatively linked to HDL and TNF-RII positively to triglyceride.
The preoperative profile with high levels of proinflammatory adipocytokines is linked to smaller improvements in glucose homeostasis and lipid factors. The use of a range of biomarkers may predict the level of metabolic changes following bariatric surgery.
Bariatric surgery is an effective treatment for morbid obesity [1, 2]. In general, patients tend to lose weight rapidly after surgery, with a negative energy balance in the first few weeks  that is linked to a significant improvement in metabolic control . Three studies, Schauer et al. , Mingrone et al. , and the most recent Carlsson et al. , have all reported the effect of bariatric surgery on metabolic control and the prevention of type 2 diabetes in obese patients. They all suggest that in certain groups of patients treatment by bariatric surgery is indicated to correct metabolic comorbidities, regardless of weight loss. As Jacobs  points out, with such a large number of people being eligible for this treatment, there is a need for a better understanding of the pathophysiological mechanisms that provoke these changes.
At present, predicting the extent of weight loss after bariatric surgery is often not very accurate, and basal BMI is still the best predictor of the expected BMI postoperatively . However, a model has not been yet developed that predicts the degree of metabolic comorbidity reversal, which, apart from weight loss, might become the goal of surgical treatment. There is, therefore, a need to develop methods for predicting the metabolic effects of bariatric surgery.
A number of studies have been developed to address the role of adipo/cytokines following bariatric surgery [9-11]. Adiponectin and leptin levels after bariatric surgery have been related to insulin resistance and weight reduction; but the use of the preoperative levels of these adipo/cytokines for predicting weight loss and the amelioration of insulin resistance, hypertension, and dyslipidemia has not been evaluated yet. In addition, the development of biomarker panels may improve the performance of the individual molecules.
Adipose tissue, apart from its classical role as an energy storage depot, is also a major endocrine organ that secretes many factors, whose local and circulating levels are affected by the degree of adiposity. In this study, we included, together with adiponectin, another adipocytokine visfatin that is related to fat mass and glucose homeostasis in obesity [12, 13]. Evidence has emerged over the last decade that shows inflammation to be critical in the development of insulin resistance, diabetes, and related diseases, and obesity is now considered to be a state of chronic low-grade inflammation. In addition to adipocytokines, we studied factors related to inflammation, such as C reactive protein (CRP), and receptors of tumor necrosis factor (TNF-RI and TNF-RII).
The aim of this study was to evaluate preoperative and postoperative serum levels of adipocytokines and their relationship with any changes in metabolic factors caused by bariatric surgery in morbidly obese women. We included an evaluation of pro-inflammatory adipocytokines, TNF-RI, TNF-RII, visfatin, and CRP, and the anti-inflammatory adipocytokine, adiponectin. All of them are related to fat mass and glucose homeostasis in obesity [13-15]. The principal objective was to elucidate whether combinations of these biomarkers could improve their predictive performance and to establish an approach for the identification of the most appropriate candidates for bariatric surgery in order to improve metabolic alterations and lower cardiovascular risk.
In this study, we analyzed the adipocytokine circulating levels in 30 morbidly obese (MO) women of European descent with a BMI > 40 kg/m2 and 60 normal weight women with a BMI< 25 kg/m2 (lean controls). Follow-up samples were obtained from morbidly obese patients who underwent one of two types of laparoscopic bariatric surgery: sleeve gastrectomy (n = 17) or roux-en-Y gastric by-pass (n = 13). MO women were studied at three time-points: before bariatric surgery started (baseline), and at 6 and 12 months after surgery. The institutional review board approved the study. All participants gave written informed consent for participation in medical research.
The weight of all subjects was stable, i.e. with no fluctuations greater than 2% of their body weight, for at least 3 months before surgery. Those patients who had an acute illness, acute or chronic inflammatory or infective diseases or end stage malignant disease were excluded from this study, as were patients with any severe complication after surgery. Menopausal women and women receiving contraceptive treatment were also excluded.
Anthropometrical evaluation included measures of BMI and waist circumference (WC). Laboratory studies included glucose, insulin, glycosylated hemoglobin A1c (HbA1c), high density lipoproteins (HDL), and triglycerides, which were carried out using a conventional automated analyzer and measured after overnight fasting. The homeostasis model assessment of insulin resistance (HOMA2-IR) was completed using the HOMA Calculator version 2.2.2  (http://www.dtu.ox.ac.uk, accessed May 2010).
We determined the circulating levels of different molecules related to inflammation including adipokines (high molecular weight (HMW) adiponectin and visfatin), acute phase proteins (CRP), and proinflammatory cytokines (TNF-RI and TNF-RII). Circulating levels of TNF-RI, TNF-RII (AssayPro, St. Charles, USA), CRP (Dade Behring, Marburg, Germany), HMW adiponectin (Millipore, Missouri, USA), and visfatin (AdipoGen Inc., Seoul, Korea) were measured in duplicate using enzyme-linked immunosorbent assays (ELISA) following the manufacturers instructions. Adiponectin assay sensitivity was 0.78 ng/ml and inter-assay and intra-assay coefficients of variation were less than 8.4 and 7.4%, respectively. CRP assay sensitivity was 0.2 ng/ml and inter-assay and intra-assay coefficients of variation were less than 4.8 and 3.8%, respectively. The TNF-RI assay sensitivity was 50 pg/ml and the inter-assay and intra-assay coefficients of variation were less than 5.7 and 1.7%, respectively. The TNF-RII assay sensitivity was 0.1 ng/ml and the inter-assay and intra-assay coefficients of variation were less than 3.2 and 3.3%, respectively. The visfatin assay sensitivity was 30 pg/ml and the inter-assay and intra-assay coefficients of variation were less than 7.2 and 9.1%, respectively.
All the values reported are expressed as mean ± standard deviation (SD) and were analyzed using the SPSS/PC+ for Windows statistical package (v.19.0 Chicago, IL). Differences between the morbidly obese and the lean controls were calculated using the Student's t test (independent or related samples) for normally distributed parametric variables. Non-parametric variables were analyzed using the Wilcoxon test for the differences between related samples, and the Mann–Whitney U test for independent samples. The strength of the association between variables was calculated using Pearson's method for parametric variables and the Spearman Rho correlation test for non-parametric contrasts. A multiple linear regression analysis with backward variable selection was carried out to identify independent predictors of BMI, HOMA2-IR, triglycerides, and HDL postoperatively. The validity of the regression model and its assumptions were assessed with the plot of residuals vs. predicted. P values < 0.05 were considered to be statistically significant.
Patients' baseline characteristics, metabolic variables, and adipocytokine levels are given in Table 1. Biochemical analyses indicated that MO patients had significantly higher WC, fasting glucose, insulin, HbA1c, HOMA2-IR, triglycerides, and lower HDL than the control group did. The levels of TNF-RI, TNF-RII, visfatin, and CRP were significantly higher in MO versus controls, and adiponectin levels were significantly lower in MO.
|Lean control (n=60)||Morbidly obese (n=30)|
|Mean (SD)||Mean (SD)||P-value|
|AGE (years)||44.8 (16.2)||47.2 (8.9)||ns|
|WEIGHT (kg)||61.9 (8.9)||120.1 (13.1)||<0.001|
|WC (cm)||80.7 (11.4)||133.3 (11.5)||<0.001|
|BMI (kg/m2)||23.8 (3.1)||46.5 (5.0)||<0.001|
|Glucose (mg/dL)||95.5 (15.1)||122.4 (34.9)||<0.001|
|Insulin (mU/L)||9.2 (6.3)||18.2 (11.0)||<0.001|
|HbA1c (%)||4.7 (0.4)||5.5 (1.2)||<0.001|
|HOMA2-IR||1.2 (0.9)||2.4 (1.5)||<0.001|
|HDL (mg/dL)||58.8 (7.9)||38.9 (6.4)||<0.001|
|TG (mg/dL)||100.8 (58.6)||166.1 (82.5)||<0.001|
|TNF-RI (ng/mL)||2.32 (0.63)||2.80 (0.72)||0.004|
|TNF-RII (ng/mL)||4.38 (1.82)||5.06 (2.05)||ns|
|CRP (ng/mL)||0.56 (1.63)||1.09 (0.80)||0.039|
|VISFATIN (ng/mL)||1.40 (0.64)||3.17 (2.47)||<0.001|
|Adiponectin HMW (ng/mL)||7695.66 (3448.56)||3283.59 (1929.93)||<0.001|
In the MO patients, we evaluated anthropometrical characteristics, metabolic variables, and adipocytokine levels at baseline, 6 and 12 months after bariatric surgery (Table 2).
|Morbid obese (n=30)|
|Baseline||6 Months||12 Months|
|Mean (SD)||Mean (SD)||P-value1||Mean (SD)||P-value2|
|AGE (years)||47.2 (8.9)|
|WEIGHT (kg)||120.1 (13.1)||90.5 (11.8)||<0.001||83.8 (13.4)||0.037|
|WC (cm)||133.3 (11.5)||101.6 (10.8)||<0.001||98.6 (11.1)||ns|
|BMI (kg/m2)||46.5 (5.0)||35.0 (3.9)||<0.001||32.5 (5.5)||0.036|
|Glucose (mg/dL)||122.4 (34.9)||86.8 (11.6)||<0.001||84.5 (12.3)||ns|
|Insulin (mU/L)||18.2 (11.0)||10.5 (5.7)||<0.001||8.4 (4.5)||ns|
|HbA1c (%)||5.5 (1.2)||4.7 (0.6)||<0.001||4.8 (0.6)||ns|
|HOMA2-IR||2.4 (1.5)||1.3 (0.7)||<0.001||1.1 (0.6)||ns|
|HDL (mg/dL)||38.9 (6.4)||48.2 (9.7)||0.004||49.8 (10.5)||ns|
|TG (mg/dL)||166.1 (82.5)||96.2 (32.6)||<0.001||92.7 (29.6)||ns|
|TNF-RI (ng/mL)||2.80 (0.72)||2.39 (0.46)||0.007||2.11 (0.40)||0.010|
|TNF-RII (ng/mL)||5.06 (2.05)||4.37 (1.26)||0.022||3.36 (1.03)||0.001|
|CRP (ng/mL)||1.09 (0.80)||0.56 (0.80)||0.010||0.21 (0.34)||ns|
|VISFATIN (ng/mL)||3.17 (2.47)||3.10 (1.71)||ns||2.36 (1.55)||0.036|
|Adiponectin HMW (ng/mL)||3283.59 (1929.93)||8585.27 (3914.14)||<0.001||10636.61 (4712.94)||ns|
After surgery, BMI, fasting glucose, insulin, HOMA2-IR, and HbA1c levels were markedly reduced postoperatively. Patients lost 24.7% (4.9) of their BMI 6 months after bariatric surgery and a 31.0% (6.8) after 12 months. The incidence of diabetes mellitus was 40% in our cohort at baseline and decreased to 3.3% after surgery. HDL levels increased moderately, while triglyceride levels were much lower. Accordingly, dyslipidemia affected 93.3% of the patients before surgery, 70% after 6 months, and 50% after 12 months. After analyzing the global effect of bariatric surgery on all the traits of the metabolic syndrome (MS), we wanted to investigate the potential effect of the type of bariatric surgery carried out on the melioration/improvement found in these subjects. There were no differences in the melioration of the BMI, insulin resistance, or dyslipidemia between the two types of bariatric surgery (data not shown).
Finally, the levels of TNF-RI, TNF-RII, visfatin, and CRP were significantly lower than its baseline levels, whereas HMW adiponectin levels were higher (Figure 1).
In order to create a predictive model, we first evaluated the potential contribution of all the adipocytokines analyzed at baseline. We found that only TNF-RII correlated with BMI at 6 and 12 months after surgery (r = 0.448, P = 0.013; r = 0.474, P = 0.008, respectively). In addition, TNF-RII correlated with the percentage of body mass index reduction at 12 months (r = −0.364, P = 0.044). The BMI values at baseline were strongly related with the BMI at 6 and 12 months (r = 0.764, P < 0.001; r = 0.622, P < 0.001, respectively). With these results, we included TNF-RII and BMI baseline values in the model. The best predictor of the BMI that is reached after surgery is the baseline value of BMI. These values explained 60% (R2 = 0.600; B coefficient = 0.680; P < 0.001) of the variability of BMI at 6 months and almost 40% (R2 = 0.387; B coefficient = 0.552; P < 0.001) of the variability of BMI at 12 months.
Firstly, we evaluated the relationship between HOMA2-IR postoperatively and the adipocytokine levels preoperatively. We found that HOMA2-IR values at 6 months after surgery were related to TNF-RI (r = 0.455, P = 0.015), and visfatin levels (r = 0.516, P = 0.005). At 12 months after surgery, HOMA2-IR was related to visfatin (r = 0.403, P = 0.037). When HOMA2-IR at 6 months was introduced as the independent variable, the baseline levels of TNF-RI and visfatin predicted 38% of its variability. We then created a model to predict HOMA2-IR 12 months after surgery. Again, visfatin was the molecule with the highest predictive power. On the other hand, the model generated when we included the percentage of BMI reduction at 12 months and visfatin predicted 42% of the variability of the insulin resistance index 12 months postoperatively.
Interestingly, postoperative HOMA2-IR was not predicted by baseline HOMA2-IR or baseline BMI in the cohort analyzed.
Relationships between the basal levels of visfatin and absolute changes of HOMA2-IR (▵HOMA2-IR) at 6 and 12 months were also examined. ▵HOMA2-IR 6 and 12 months were again independently associated with visfatin.
Secondly, we evaluated the relationship between insulin levels postoperatively and the adipocytokine levels preoperatively. We found that insulin levels at 6 months after surgery were related to TNF-RI (r = 0.441, P = 0.019) and visfatin levels (r = 0.503, P = 0.006). At 12 months, insulin related positively to visfatin (r = 0.394, P = 0.038) and negatively to HMW adiponectin (r = −0.407, P = 0.048). The preoperative levels of TNF-RI and visfatin (R2 = 0.363) were a predictor of insulin levels 6 months after surgery. When we created a model to predict insulin levels at 12 months, we found that HMW adiponectin and visfatin significantly predicted 30% of its variability.
Next, we found that glucose levels at 6 months after surgery were related to TNF-RI (r = 0.458, P = 0.008) and HMW adiponectin (r = −0.356, P = 0.044), and visfatin levels (r = 0.418, P = 0.017). At 12 months, glucose related positively to TNF-RI (r = 0.364, P = 0.044). The preoperative levels of HMW adiponectin and visfatin (R2 = 0.355) were a predictor of glucose levels 6 months after surgery. However, when we created a model to predict glucose levels at 12 months, we found that only TNF-RI significantly predicted 13% of its variability.
Finally, we evaluated the relationship between HbA1c levels postoperatively and the adipocytokine levels preoperatively. We found that HbA1c at 6 months after surgery related to TNF-RI (r = 0.609, P = 0.001) and HMW adiponectin levels (r = −0.441, P = 0.024). At 12 months after surgery, HbA1c related positively to visfatin (r = 0.381, P = 0.035) and with TNF-RI (r = 0.641, P < 0.001). The models indicate that the preoperative levels of TNF-RI (R2 = 0.363 and R2 = 0.441, respectively) predicted HbA1c levels at 6 and 12 months after surgery, even after adjustment for BMI and age.
None of the adipocytokines tested were related to the triglycerides or HDL-cholesterol levels after 6 months. In contrast, at 12 months after surgery, triglycerides related to both TNF-RI and TNFRII (r = 0.451, P = 0.011; r = 0.580, P = 0.001, respectively), and HDL to CRP levels (r = −0.369, P = 0.041). The predictive power of these variables is also described in Table 3.
|Independent variable||Predictive variables||R2||B coeficient||r||P-value|
|Glucose levels 6m|
|Model 1||Adiponectin HMW||0.355||−0.002||−0.370||0.042|
|Glucose levels 12m|
|Insulin levels 6m|
|Insulin levels 12m|
|Model 1||Adiponectin HMW||0.295||−0.001||−0.404||0.050|
|Model 2||% of BMI reduction 12m||0.424||-0.037||-0.507||0.004|
|HbA1c levels 6m|
|HbA1c levels 12m|
In this study, we found interesting changes in the levels of adipocytokines and metabolic factors after bariatric surgery. The levels of the pro-inflammatory mediators TNF-RI, TNF-RII, and CRP decrease following surgery, whereas anti-inflammatory adiponectin levels increase. These results agree with previous reports where body weight reduction also exerts potent inhibitory effects on pro-inflammatory activity in adipose tissues [3, 9-11]. All the changes found in adipocytokine levels are statistically significant at 6 months after surgery, except visfatin levels that are significantly lower at 12 months.
In our cohort of morbidly obese women, in addition to bariatric surgery inducing weight loss, the metabolic factors were also modified in a way beneficial for health. The mean fasting glucose and insulin levels were markedly reduced postoperatively leading to a marked improvement in insulin sensitivity expressed as the HOMA index. Also, the levels of glycosylated hemoglobin A1c were lower postoperatively. All these findings agree with previous studies in obesity with type 2 diabetes [5, 6]. After surgery, triglycerides were lower and HDL-C was higher.
We evaluated the potential predictive power of preoperative adipocytokine levels in relation to weight loss, and changes in the metabolic factors at 6 and 12 months postoperatively. Our results indicate that the best variable for predicting the achievable loss of weight after bariatric surgery is the preoperative BMI value. These results support the notion that the more obese patients tend to lose more weight [3, 17]. However, the baseline BMI is not a significant predictor of the reversal of the obesity metabolic comorbidities. Interestingly, we found that higher basal pro-inflammatory adipocytokine levels relate to a minor decrease in insulin resistance. Specifically, higher basal levels of TNF-RI and visfatin predict a lower reduction of HOMA values at 6 months postoperatively. Visfatin predicts changes of the HOMA2-IR at 12 months independently of the baseline BMI, where higher concentrations of visfatin at baseline are associated with smaller declines in HOMA2-IR values.
High baseline TNF-RI and visfatin levels are predictors of minor reductions in insulin levels after surgery. Patients with higher preoperative levels of HMW adiponectin reduced glucose at 6 months and insulin at 12 months by a greater amount. Higher visfatin levels at baseline also relate to a smaller reduction in glucose and insulin levels after surgery. A high level of TNF-RI predicts a smaller decrease in glucose and HbA1c levels at 12 months postoperatively, even after adjustment for BMI and age.
In summary, high levels of the pro-inflammatory adipocytokines TNF-RI and visfatin are associated with minor reductions in glucose, insulin, HbA1c, and HOMA2-IR postoperatively. However, the anti-inflammatory molecule adiponectin positively relates to a major amelioration of these parameters, which agrees with a previously described role of adipocytokines in insulin resistance [18, 19].
Regarding lipid parameters, HDL levels increase significantly after surgery while triglyceride levels sharply decrease, underlining the beneficial effects of weight loss on major cardiovascular risk factors [2, 20, 21]. The changes in triglycerides and HDL values postoperatively correlate with pro-inflammatory factors; high CRP basal levels therefore correlate with a smaller increase in HDL and TNF-RII correlates with a smaller decrease in triglyceride values at 12 months. In the same way as we found with glucose metabolism, higher values of pro-inflammatory factors predict a lower response in the degree of amelioration of lipid parameters. Finally, preoperative levels of HMW adiponectin have been described to independently predict weight loss following a gastric by-pass  but in this study we did not observe any significant predictive power of this molecule on weight loss or changes in lipid parameters.
The outstanding finding of our study has been the relationship found between high preoperative subclinical inflammation and limited metabolic improvement following bariatric surgery. These results suggest that a preoperative inflammatory profile with high levels of pro-inflammatory adipocytokines and low values of adiponectin are associated to smaller improvements in biochemical-metabolic factors of glucose homeostasis and lipid profile in morbidly obese women at 12 months after surgery. In obesity, adipose tissue expansion is associated with local infiltration of different types of inflammatory cells . These cells secrete different cytokines, which, in turn, alter the expression and secretion pattern of adipokines and cytokines in adipose tissue. Accordingly, it has been demonstrated that insulin-resistant morbidly obese individuals have a significantly higher number of macrophage infiltration in their omental adipose tissue compared with insulin-sensitive individuals . In this scenario, adipose tissue inflammation is thought to play a role in the induction of adipose tissue dysfunction. In our cohort, it seems reasonable that the adipose tissue dysfunction might be higher in patients with higher subclinical inflammation, which might complicate the metabolic improvement.
These preliminary results were obtained in a homogeneous cohort of morbidly obese women after a follow-up period of 12 months. The main limitation of the study is that our results are not extrapolable to other populations. The predictive power of adipocytokines and other biomarkers needs to be addressed in a larger independent cohort including different genders, degrees of obesity, and longer follow-up periods.
These findings need to be verified and this approach is justified in order to investigate a range of biomarkers that, associated to clinical data, could predict the level of metabolic changes following bariatric surgery. This data would allow specialists to establish some criteria for the selection of obese patients with metabolic comorbidities for whom bariatric surgery would have the greatest benefit. Further studies are needed in order to establish a panel of cut-off values that would define the predictive model.