Improved Arteriole-to-Venule Ratio of Retinal Vessels Resulting From Bariatric Surgery




Obesity causes increased morbidity and mortality from metabolic and cardiovascular disease (CVD). We investigated the effect of bariatric surgery on endothelial dysfunction (ED) in retinal vessels as a marker of metabolic and cardiovascular risk in patients with obesity WHO III. Thirty consecutive patients (19/11, w/m) were evaluated by anthropometry, lipid profile, and oral glucose tolerance test before and after bariatric surgery (Mannheim Obesity Study (MOS); NCT 00770276). Risk stratification was performed by the presence of metabolic syndrome (MetS) according to ATP-III (adult treatment panel-III). Subclinical atherosclerosis was assessed by measurement of intima-media thickness (IMT). Flicker light response of retinal vessels was used as measures of ED. We measured their arteriole-to-venule ratio (AVR) for evaluation of vascular pathology. After a median of 9 months following bariatric surgery, mean weight loss was 39.4 kg (37.3%). Remission of impaired glucose metabolism was achieved in 53.3% of affected patients. Dyslipidemia improved significantly (triglycerides −61.3 mg/dl, P < 0.0001, total cholesterol −28.2 mg/dl, P = 0.002, and low-density lipoprotein cholesterol were reduced −24.5 mg/dl, P = 0.008). This resulted in a significant reduction of patients classified for MetS (27 vs. 9, P < 0.0001). Adiponectin increased by 2.08 µg/l (P = 0.032) and high sensitivity C-reactive protein (hs-CRP) and soluble intercellular cell adhesion molecule (sICAM) decreased (−7.3 mg/l, P < 0.0001 and −146.4 ng/ml, P = 0.0006). AVR improved significantly (+0.04, P < 0.0001), but neither Flicker light response nor IMT changed significantly. Retinal AVR is ameliorated after bariatric intervention. As an increased AVR results from either or both widening retinal arteriolar caliber and narrowing retinal venular caliber, an improvement in small vessel profile is evident 9 months after bariatric surgery.


Obesity is an established risk factor for the development of diabetes mellitus and dyslipidemia resulting in accelerated atherosclerosis and cardiovascular disease (CVD) (1). Bariatric surgery is the most effective treatment that favorably reduces metabolic sequelae of abdominal obesity (2). Consequently, observational and interventional studies have shown reduced cardiovascular morbidity and mortality in response to surgically induced weight loss (3,4). Reduction of specific metabolic abnormalities evolves within weeks or months after bariatric surgery (5), even before weight reduction is established. In contrast, the benefit on cardiovascular morbidity and mortality takes years to develop (3,4). Bridging this gap, clinical studies have used surrogate markers of endothelial function such as arterial flow-mediated vasodilatation (FMD) (6).

CVD is preceded by endothelial dysfunction (ED) and subclinical atherosclerosis (7). Apart from, and mediated through classical risk factors such as smoking and hypercholesterolemia, changes in the vascular endothelium with low-grade inflammatory processes lead to a continuum of vascular damage including expression and secretion of vascular adhesion molecules, impaired nitric oxide-mediated vasodilatation, subendothelial accumulation of modified and unmodified lipids with infiltration of mononuclear cells, and others. Inflammatory cytokines can mediate oxidative stress and vascular toxicity (7). The adipose tissue has been identified as an endocrine organ (8). It is a primary source of cytokines among other factors promoting premature atherosclerosis (8). Therefore, obesity may lead to premature atherosclerosis and vascular disease.

Methods used to establish ED and premature atherosclerosis in clinical studies include ultrasound-based measurement of intima-media thickness (IMT), vasodilatation of blood vessels of the forearm, and measurement of soluble vascular adhesion molecules. All of them, however, represent a different aspect of the atherosclerotic process (9,10,11).

Starting with work by Robert Marcus Gunn, it has long been appreciated that ED can also be assessed by inspection of retinal vessels (Gunn's sign = arteriovenous nicking). Changes induced by cardiovascular risk factors, obesity, the metabolic syndrome, and diabetes, include structural lesions such as microaneurysms, dot blots, or arteriovenous nicking, as well as static and dynamic alterations of the structurally intact vessel system. The arteriole-to-venule ratio (AVR) of retinal vessels as the most important static parameter of endothelial function is associated with CVD, independent of classic cardiovascular risk factors (12,13). Changes in AVR can result from generalized arteriolar narrowing, or from generalized venular dilatation, or both. For example, arteriolar narrowing indicates chronic vasoconstriction from elevated blood pressure (14,15,16). Increased venular diameters are associated with obesity, the metabolic syndrome, systemic inflammation, and dyslipidemia (17,18,19). Therefore, the AVR reflects metabolic and cardiovascular risk translated into ED. Dynamic ED can be assessed by vasodilatation of retinal vessels through flicker light stimulation (20,21,22). The neurovascular unit of the retina is thought to provide vasodilatory nitric oxide from astrocytes when stimulated by flickering light, leading to dilatation of retinal arteries and veins (23). Impaired flicker response has been observed in diverse conditions associated with ED such as diabetes, arterial hypertension, and hyperlipidemia (21,24,25). A modest, but significant correlation of retinal ED with FMD was established in patients with ED and healthy controls (26).

Given the paucity of data on ED by retinal vessel assessment, and the lack of interventional data on AVR, we aimed at studying the effect of bariatric surgery on retinal vessel as an early surrogate marker of atherosclerosis in patients with obesity WHO III undergoing bariatric surgery. Here, we provide first evidence that weight loss from bariatric surgery results in improvement of AVR.

Methods and Procedures

We conducted a prospective study in obesity WHO III with consecutive enrollment (clinical trials: NCT00770276). Patients were examined before and after bariatric surgery for metabolic and vascular abnormalities. The study protocol was approved by the local ethics committee. This study summarizes data of the first 30 patients (w/m 19/11). This first analysis included 23 patients after Roux-en-Y gastric bypass (GBP), five sleeve gastrectomies, one laparoscopic gastric binding, and one biliopancreatic diversion.

Medical history was evaluated by a standardized protocol. We examined for cardiovascular risk factors (preexisting CVD, family history of CVD, smoking, and arterial hypertension). Patients were interviewed for associated diseases and medication. Height and body weight were determined by calibrated scales to the closest of 0.1 kg and cm, respectively. Further anthropometry included neck circumference as an estimate of subcutaneous fat tissue, arm and waist circumference, and of blood pressure. Metabolic analysis was performed after an overnight fast. Baseline measurements included blood glucose, serum insulin, cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and triglycerides. Adiponectin (human adiponectin ELISA; Mediagnost, Reutlingen, Germany), sICAM (Diaclone, Cologne, Germany), sVCAM (R&D Systems, Wiesbaden, Germany), hs-CRP (Dimension RxL; Siemens, Eschborn, Germany), IL-6 (R&D Systems), TNF-α (Quantikine human TNF-α/TNFSF1A Immunoassay; R&D Systems), and MCP 1 (Diaclone) were measured by commercial available ELISAs. We screened for the presence of impaired glucose metabolism by 75 g oral glucose tolerance test according to WHO/International Diabetes Federation (IDF) guidelines. Patients were stratified for MetS according to ATP-III (adult treatment panel-III) definition (27). After individual calculating of the PROCAM (Prospective Cardiovascular Münster) and Framingham risk score, we classified the study group according to CVD risk defined by NCEP (National Cholesterol Education Program)-categories (low: 1, 0–1 risk factor; moderate: 2, two or more risk factors; and high: 3, CVD or CVD equivalent (diabetes mellitus, Framingham risk score >20%)) (28,29).

Measurement of IMT of the right carotid artery was assessed to identify the preexisting risk of atherosclerosis. We followed the imaging protocol of the consensus statement of the American Society of Echocardiography (9).

Assessment of ED—AVR and flicker reaction

Digital video analysis was used for digital fundus imaging for conventional examinations and for retinal vessel analysis (RVA; Imedos, Jena, Germany). After mydriasis with 10% phenylephrine and 1% tropicamide, pictures centred for the macula, the inferior arcade and optic disc were recorded from both eyes. AVR was calculated by the software after using arteries and veins within two diameters of the optic disc (RVA; Imedos). AVR was assessed in a blinded fashion. Dynamic ED was measured by flicker reaction of retinal vessels. The method uses records of vessel dilatation in response to flickering light (20). In detail, diameters of a vein and artery were continuously measured using 1.5 mm long segments of vessels within two disc diameters of the center, which showed no crossing, bifurcation or tortuosity of >30°. After baseline calibration during an initial 50-s period, flickering light was applied for 20 s by interruption of continuous light with 12.5 Hz. During this and a subsequent period of 80 s, vessel diameters were continuously recorded. For improved accuracy, the procedure was repeated three times. Dilatory responses of the artery and vein were expressed as the maximum dilatation relative to the baseline diameter. Results were compared with data from the ARIC (Atherosclerosis Risk in Communities) study (30).

Statistical analysis

Data are expressed as means ± s.d. or as total number/percent (except for gender distribution w/m). The significance of differences in means was calculated using Wilcoxon-rank test. Categorical data was analyzed by χ2 test. A P value of <0.05 was considered significant. All statistical analyses were performed using the statistical software StatsDirect for Windows (Version 2.7.0; Statsdirect, Altrincham, UK).


After a median follow-up of 9 months (range 6–23 months), we found a significant reduction of body weight of 39.4 kg (−27%) (Table 1), which resulted in a significant decrease of BMI (−13.4 kg/m2). Four patients achieved a BMI below 30 kg/m2. We observed a reduction of diastolic blood pressure (−7.2 mm Hg; −8%; P = 0.006) as well as of neck (−5.9 cm; −13%; P < 0.0001) and waist circumference for the whole cohort (−24.3 cm, −18%; P < 0.0001), respectively. Time from operation did not correlate with changes in weight loss or BMI (P = 0.31 and P = 0.22, respectively).

Table 1.  Anthropometrical data of patients before and after bariatric surgery (n = 30, f/m = 19/11)
inline image

Bariatric surgery resulted in highly significant improvements of glucose metabolism with a mean reduction of fasting glucose (−11.1 mg/dl; −12%; P < 0.0001), 2 h oral glucose tolerance test glucose (−49.7 mg/dl; −38%; P < 0.0001), serum insulin (−18.5 mU/l; −68%; P < 0.0001), and homeostasis model assessment of insulin resistance (−7.1; −80%; P < 0.0001). In addition, we observed a highly significant, albeit quantitatively modest decline in total (−28.2 mg/dl; −13%; P = 0.017) and low-density lipoprotein cholesterol (−24.5 mg/dl; −17%; P = 0.008) and a substantial decrease in triglyceride levels (−61.3 mg/dl; −34%; P < 0.0001) (Table 2). Interestingly, high-density lipoprotein increased only modestly after bariatric surgery, not quite reaching statistical significance (+3.4 mg/dl, P = 0.059).

Table 2.  Adipokines, soluble markers of endothelial dysfunction, markers of inflammation, and glucose as well as lipid metabolism before and after bariatric surgery
inline image

As expected, bariatric surgery resulted in significant remission of impaired glucose tolerance and type 2 diabetes (8 vs. 1, P = 0.03, 11 vs. 1, P = 0.03, respectively; Table 3). Impaired glucose metabolism persisted only in patients with preexisting diabetes. Amelioration of glucose metabolism together with improved blood pressure and lipid profiles resulted in reduced prevalence of the MetS (26 vs. 9, P < 0.0001).

Table 3.  Risk factors and risk stratifications according to metabolic and cardiovascular risk before and after bariatric surgery
inline image

Next, we assessed blood-borne parameters of ED, and adipokines with known effects on vascular function in obesity. Vasoprotective adiponectin demonstrated a significant rise in both genders (mean rise 1.89 µg/l, P = 0.01). A 19.1% reduction was found for sICAM-1 (−146.4 mg/l, P = 0.0006). In contrast, soluble vascular cell adhesion molecule-1 (sVCAM-1) was not reduced through metabolic improvement. Hs-CRP as a marker of inflammation was reduced after bariatric surgery (−7.3 mg/l, −57%; P < 0.0001). Of particular note, IL-6, TNF-α, and MCP-1 did not show any significant change.

Bariatric surgery resulted in an amelioration of endothelial function. AVR increased significantly in both genders (+0.04, P < 0.0001) (Table 4, Figure 1). Time from operation, reduction of anthropometric parameters, changes of metabolism, adipokines, cytokines, or markers of inflammation were not correlated with the change in AVR (data not shown). The increase of AVR was the result of both, a significant dilatation of retinal arteries (+4.28, P = 0.025) and a decrease of retinal vein calibers (−5.73, P = 0.012). Flicker light-induced dilatation of retinal veins showed a trend for increased reactivity (+1.25%, P = 0.061), whereas flicker light-induced dilatation of arteries remained unchanged. Subclinical atherosclerosis as measured by IMT was reduced by 0.04 mm, not yet reaching significance level (P = 0.15).

Table 4.  Endothelial function and subclinical atherosclerosis (IMT) before and after bariatric surgery
inline image
Figure 1.

Improvement of endothelial function after bariatric surgery. Arteriole-to-venule ratio of retinal vessels is displayed before and after bariatric surgery.

Finally, when assessing cardiovascular risk stratified by NCEP categories there was a significant reduction in the overall CVD risk after surgery. This was mainly a result of remission of diabetes mellitus II as a CVD equivalent (Tables 2 and 3). Amelioration of cardiovascular risk was observed with a 4.27% reduction of the PROCAM risk score (P = 0.01). This was achieved through surgery, and not through intensified medication, because fewer patients required treatment for dyslipidemia, hypertension or impaired glucose metabolism (remission in 2 of 5 = 40%, 3 of 11 = 27.3%, and 6 of 7 = 85.7%, respectively).

Due to the different surgical techniques, we analyzed the subgroup which underwent Roux-en-Y GBP (n = 23) separately. In this sample, we found similar amelioration of anthropometrical and metabolic parameters as in the entire cohort (Table 1 in Supplementary Appendix online). However, a significant increase of high-density lipoprotein cholesterol and adiponectin as well as a significant decrease of hs-CRP were noted after GBP. sICAM as a soluble marker of ED was also significantly reduced. Flicker reaction of retinal veins improved significantly after GBP (+1.3%, P = 0.032) (Table 5). Arterial reaction after flicker light showed a trend for reduced reaction (−0.63%, P = 0.092). As in the entire cohort, AVR improved significantly after surgical weight management (+0.04, P < 0.0001). Correlation analysis of flicker light reaction of retinal vessels and AVR did not show any association with time from operation or weight loss.

Table 5.  Endothelial function and subclinical atherosclerosis (IMT) before and after Roux-Y gastric bypass
inline image


This study shows that bariatric surgery improves AVR of retinal vessels in patients with obesity WHO III in parallel with improvements of parameters of lipid and glucose metabolism as well as amelioration of adipocytokines and markers of inflammation involved in premature atherosclerosis. Improved retinal vessel AVR resulted from both, corrections of constricted retinal arteries and inadequate dilatations of veins. Furthermore, the patients after GBP demonstrated improved endothelial function of retinal veins as measured by flicker reaction.

The association of impaired metabolism, obesity, and changes of retinal vessel has been determined in large studies such as ARIC and Funagata study. Although assessed in different genetic backgrounds, there were remarkable consistencies found in both studies. Generalized arteriolar narrowing reflected elevated blood pressure, and dilatation of venular diameter reflected obesity, hyperglycemia, systemic inflammation, and dyslipidemia (14,15,16,17,18,19,31). Therefore, the increase of AVR likely results from the beneficial changes of these parameters by bariatric surgery. Retinal AVR and brachial FMD, although only modestly correlating, are predictive of cardiovascular health (12,13,26,32). FMD as a surrogate measure of ED has been demonstrated to improve after bariatric surgery. Habib et al. demonstrated increased FMD established at a median of 6 months after Roux-en-Y bypass (6). Gokce et al. were also able to present superiority of Roux-en-Y GBP over conservative weight reduction with regard to improvement of FMD (33). Interestingly enough, the effect was associated with glycemia, but not with blood pressure, dyslipidemia, and the degree of weight loss. Though we found no association of changes in metabolism and AVR, comparatively large arterial vessels of the forearm (FMD) seem to respond differently to amelioration of metabolism after bariatric surgery. Large cross-sectional studies have documented an association of AVR to CVD independent of classical cardiovascular risk factors (12,13). Wang et al. determined a 1.5-fold higher risk of coronary heart disease death for every standard deviation reduction of AVR (34). Consequently, an improvement of AVR due to therapeutic intervention could reflect a reduction of cardiovascular risk. Of note, to the best of our knowledge, this is the first study to demonstrate an improvement in AVR of retinal vessels following intervention. Moreover, ED measured by flicker reaction of retinal veins decreased after GBP. Perhaps the more complex intervention in regard to gastric sleeve resection resulting in different regulation of intestinal hormones and absorption may be causative for this finding. Interestingly, time from operation, changes of anthropometry and metabolism, cytokines, adipokines or markers of inflammation showed no significant association with changes of AVR or flicker light reaction. These conflicting findings may be explained by two speculations. First, the improvement of small vessel dysfunction is a consequence of the cumulative effect of all parameters affected by bariatric surgery. Alternatively, an early change due to the missing link to time from operation of different factors (e.g., incretins) is more or also relevant. With our data, however, we cannot clearly decide which hypothesis is applicable. Moreover, we were unable to demonstrate an improvement of flicker reaction of retinal arteries in our entire cohort. Probably, the only modest effect on systolic blood pressure was also causative for this finding. Because studies on this issue are not available, this has to be addressed in the future.

We also studied circulating markers of ED and adipokines to disclose factors associated with improved endothelial function after bariatric surgery. Adiponectin, a vasoprotective adipokine causing an increase in endothelial nitric oxide production and modulation of expression of adhesion molecules, was ameliorated in our cohort (35). This finding is consistent with a study by de la Torre after bariatric surgery (36) who observed a similar increase of adiponectin (23 vs. 29% in our study) in females after 9 months. Soluble forms of cellular adhesion molecules are found in the circulation, and the current evidence suggests that they reflect early ED in patients particularly prone to accelerated atherosclerosis. The native forms such as ICAM-1, VCAM-1, and l-selectin are found in atherosclerotic plaques, and prospective studies have identified a predictive role of sICAM-1 in initially healthy people, and of sVCAM-1 in patients at high risk or with overt CVD (11). The significant decrease of sICAM-1, found in our study, is in accordance with published data (37). Konukoglu detected a reduction of sICAM as early as 1 month after bariatric surgery (38). Moreover, the later study did not reveal any change in levels of sVCAM-1, which complies with our findings and would indicate that our study population is positioned early in the natural course of atherosclerosis, in particular, because of female preponderance. Adipose tissue elaborates proinflammatory cytokines such as TNF-α, and propagates subinflammatory vascular response via CRP. Of several inflammatory markers measured in our study, only hs-CRP was significantly reduced, whereas IL-6 and TNF-α was unchanged at the time points of follow-up after bariatric surgery. This type of data answers important clinical questions, i.e., whether the effect of bariatric surgery on inflammation is lasting, and whether it affects all components of the obesity-induced inflammation. Similar results were observed in a small study of eight patients, however, undergoing gastric banding (39). The increase of adiponectin as a vasoprotective adipokine, reduction of sICAM, which predicts the risk for CVD and diabetes mellitus (11), as well as the reduction of hs-CRP surely amplify and or translate the beneficial effects of weight reduction on ED.

CVD risk assessments and risk engines demonstrate a beneficial effect of bariatric surgery in our population. Our study data match with a large meta-analysis by Batsis et al. (40), in which a reduction of the 10-year CVD risk by treatment was found from 7 to 3.5%. Thus, our study cohort is valid in terms of comparability in the reduction of CVD risk, and thus is informative towards the interpretation of therapeutic changes in AVR translating into a reduction in total CVD risk.

The major limitation of our study is the small sample size. Furthermore, the time to first postoperative visit was variable, and comparatively long. However, we did not find any association between time from operation and changes in AVR, suggesting that our data and conclusions are robust. Support for that view comes from the study by Habib et al. who found that ED improved within 6 months after surgery with no further change thereafter (6). Other limitations relate to changes in eating behavior and physical activity which may have also influenced our findings.

In conclusion, our study shows that retinal AVR improve after bariatric surgery. As AVR is a marker representing macrovascular mortality (12,13), hypertension, inflammation as well as impaired metabolism (14,15,16,17,18,19), we hypothesize that the increase of AVR is the result of bariatric surgery on these risk factors. ED also improved in patients after GBP perhaps presenting a superiority of this procedure as compared to only restrictive techniques, but this has to be addressed in future studies.


Supplementary material is linked to the online version of the paper at http:www.nature.comoby


The authors declared no conflict of interest.

See the online ICMJE Conflict of Interest Forms for this article.