Abstract
Objective. In nondiabetic healthy individuals, insulin secretion and sensitivity are linked by a negative feedback loop characterized by a hyperbolic function. We aimed to study the association of traditional insulin resistance (IR) factors with insulin secretion and sensitivity, and to determine whether the hyperbolic equilibrium of this relation is preserved in patients with rheumatoid arthritis (RA).
Methods. This was a cross-sectional study encompassing 361 nondiabetic individuals: 151 with RA and 210 controls. Insulin, C-peptide, and IR indices by homeostatic model (HOMA2) were assessed. A multivariable analysis was performed to evaluate the differences in the correlation of traditional IR-related factors with glucose homeostasis molecules, as well as IR indices between patients and controls. Nonlinear regression analysis was used to assess the hyperbolic relation of insulin sensitivity and secretion.
Results. HOMA2-IR indices were higher in patients with RA than controls. Hepatic insulin extraction, as assessed by the insulin:C-peptide molar ratio, was lower in patients with RA after multivariable analysis (0.08 ± 0.02 vs 0.14 ± 0.07, p < 0.001). Traditional IR-related factors showed significantly lower adjusted correlation coefficients with IR indices in patients with RA. The association between insulin sensitivity and secretion showed a different hyperbolic relation in patients with RA: the variability explained by the curve was lower in RA (nonlinear r2 = 0.845 vs r2 = 0.928, p = 0.001) and β coefficients (−0.74, 95% CI −0.77 to −0.70 vs −1.09, 95% CI −1.17 to −1.02, ng/ml, p < 0.001) were different in RA.
Conclusion. The traditional factors associated with IR in healthy individuals are less related to IR in patients with RA. Insulin sensitivity and secretion yield a different hyperbolic equilibrium in RA.
The relationship between glucose and insulin in the basal state reflects the balance between hepatic glucose output and insulin secretion, which is maintained by a feedback loop between the liver and β-cells. It has been postulated1,2 that this relationship is hyperbolic (i.e., the product of insulin sensitivity and insulin secretion is a constant). Therefore, insulin sensitivity and insulin secretion are linked by a negative feedback loop. In this regard, when an individual becomes more insulin resistant, the β-cell function is enhanced. A likely consequence of this interaction is that the magnitude of the change in the β-cell response accompanying a change in insulin sensitivity would depend on the initial degree of insulin sensitivity. Thus, in subjects with marked insulin resistance (IR), additional small changes in insulin sensitivity would produce large changes in insulin levels, whereas in very sensitive subjects, large changes in sensitivity would be associated with small changes in insulin concentrations. In the progression from normal glucose tolerance to impaired glucose tolerance, this compensatory mechanism seems to weaken over the short term until eventually deteriorating, thereby compromising β-cell compensation for ambient insulin resistance, leading to overt β-cell failure, and finally, the onset of type 2 diabetes3.
The elevation in basal insulin levels is a well-documented manifestation of IR. The condition exists when insulin levels are higher than expected relative to the level of glucose4. Although some reports have not found this association5,6, an increased prevalence of IR has been observed in patients with rheumatoid arthritis (RA)7,8,9,10,11, whereby high-grade systemic inflammation is thought to be implicated in the development of IR. However, although the phenomenon of compensatory hyperinsulinemia is known to exist in RA, the quantitative nature of the relationship between insulin sensitivity and insulin secretion in subjects with RA has not been studied, to our knowledge. To accomplish these goals, we analyzed data from a large number of patients with RA and healthy subjects. We hypothesized that in RA, a recognized IR state, the relationship between insulin secretion and sensitivity would be manifestly different compared to healthy subjects.
MATERIALS AND METHODS
Study participants
This was a cross-sectional study that included 361 white nondiabetic individuals, 151 patients with RA, and 210 controls. All patients with RA were at least 18 years old and fulfilled the 2010 American College of Rheumatology/European League Against Rheumatism diagnostic criteria12. They had been diagnosed by rheumatologists and were periodically followed up at rheumatology outpatient clinics. For the purpose of inclusion in our present study, RA disease duration was required to be ≥ 1 year. Although anti–tumor necrosis (TNF)-α treatment has been associated with changes in insulin resistance7,13,14,15, patients with RA undergoing TNF-α antagonist therapy were not excluded from our study. The control group consisted of patients recruited from the Spanish Camargo Cohort16,17. This cohort spanned the time from February 2006 to February 2011, and enrolled individuals have been followed ever since. The aim of our study was to evaluate the prevalence and incidence of metabolic bone diseases and mineral metabolism disorders. Patients and controls with diabetes mellitus were not included. Therefore, none of the patients or controls was receiving glucose-lowering drugs or insulin therapy. All patients and controls had a glycemia level < 7 mmol/l. Patients and controls were excluded if they had a history of myocardial infarction, angina, stroke, a glomerular filtration rate < 60 ml/min/1.73 m2, a history of cancer, or any other chronic diseases or evidence of infection. None of the controls was receiving glucocorticoids. However, because they are often used in the management of RA, patients taking prednisone, or an equivalent dose ≤ 10 mg/day, were not excluded. The study protocol was approved by the institutional review committee at Hospital Universitario de Canarias and Hospital Universitario Marqués de Valdecilla (both in Spain), and all subjects provided written informed consent (IRB approval number: 2015-34).
Data collection
Surveys of patients with RA and controls were performed in the same way. Subjects completed a cardiovascular risk factor and medication use questionnaire and underwent a physical examination to determine their anthropometric measures and blood pressure. Medical records were reviewed to ascertain specific diagnoses and medications. Waist circumference was measured at the intermediate circumference between the rib cage and the iliac crest while the subject was in a standing position. Hypertension (HTN) was defined as a systolic or a diastolic blood pressure < 140 and 90 mmHg, respectively. Dyslipidemia was defined if one of the following was present: total cholesterol > 200 mg/dl, triglyceride > 150 mg/dl, high-density lipoprotein (HDL) cholesterol < 40 in men or < 50 mg/dl in women, or low-density lipoprotein cholesterol > 130 mg/dl. In patients with RA, disease activity was measured using the 28-joint count Disease Activity Score (DAS28)18, while disease disability was determined using the Health Assessment Questionnaire19. In addition, Clinical Disease Activity Index20 and the Simple Disease Activity Index21 scores for RA disease activity were also performed as previously described.
Assessments
The homeostatic model assessment (HOMA) was performed to determine IR; specifically, in this study we used HOMA2: the updated computer HOMA model22,23. Briefly, this method consists of a structural computer model of the glucose-insulin feedback system in a homeostatic (overnight-fasted) state. The model was composed of a number of nonlinear empirical equations (one that precluded an exact algebraic solution), which describe the functions of organs and tissues involved in glucose regulation. This model can be used to determine insulin sensitivity (%S) and β-cell function (%β) from paired fasting plasma glucose and specific insulin, or C-peptide concentrations across a range of 1–2200 pmol/l for insulin and 1–25 mmol/l for glucose. In our study, we used C-peptide to calculate β-cell function, because the former is a marker of secretion. In addition, we used insulin data to calculate %S (because HOMA-%S is derived from glucose disposal as a function of insulin concentration). This computer model provides an insulin sensitivity value expressed as HOMA2-%S (where 100% is normal). HOMA2-IR (insulin resistance index) is simply the reciprocal of %S.
Insulin (Architect Abbott, 2000I) and C-peptide (Immulite 2000, Siemens) were determined by chemiluminescent immunometric assays in serum. Standard techniques were used to measure in-serum plasma glucose, C-reactive protein (CRP), Westergren erythrocyte sedimentation rate (ESR), and lipids. Blood collected from all the participants by means of venipuncture was stored at 4°C for < 4 h, centrifuged, and subsequently serum was removed and stored at −80°C.
Statistical analysis
Demographic and clinical characteristics between patients with RA and controls were compared using chi-square tests for categorical variables or Student t test for continuous variables (data expressed as the mean ± SD). For non-normally distributed continuous variables, either a Mann-Whitney U test was performed or a logarithmic transformation was made and data were expressed as a median (interquartile range). Differences in glucose homeostasis metabolism molecules and HOMA indices were studied by 3 different multivariable linear regression models: a univariate unadjusted model; a second model adjusted for those variables with a p value < 0.20 in differences between patients and controls (age, sex, waist circumference, dyslipidemia, statins, antihypertensive treatment, and CRP and cholesterol levels); and third, a model adding glucocorticoid intake as a binary variable and adjusted for the same variables described before. Additionally, to further control potential confounding effects, we obtained a propensity score using as independent variables age, sex, waist circumference, dyslipidemia, statins, antihypertensive treatment, CRP and cholesterol levels, and glucocorticoid intake; and group (control vs patients with RA) as dependent variable. Residuals distribution was tested and normality was confirmed in all the multivariate models. Correlations between variables were studied through Pearson’s correlation coefficient. Fisher r-to-z transformation was assessed to study the differences in Pearson’s correlation coefficients between patients and controls. Nonlinear regression was used to study the relationship between insulin sensitivity (HOMA2-%S) and insulin secretion (C-peptide), also adjusted for age, sex, waist circumference, dyslipidemia, statins, antihypertensive treatment, and CRP and cholesterol levels, and glucocorticoid intake. For all analyses, we used a 5% two-sided significance level, and all analyses were performed using IBM SPSS Statistics version 21 software (IBM), STATA version 13/SE software (StataCorp), and Statgraphics Centurion XVII version 17.2.0 (StatPoint Technologies). A p value < 0.05 was considered statistically significant.
RESULTS
Demographic, laboratory, and disease-related data
A total of 361 participants (151 patients with RA and 210 controls), with a mean (± SD) age of 53 ± 11 and 58 ± 9 years, respectively, were included in our study. The demographic and disease-related characteristics of the participants are shown in Table 1. There were no differences between patients and controls regarding body mass index (BMI). However, waist circumference was found to be higher in patients compared to controls (96 ± 13 vs 92 ± 14, p = 0.015). Although the presence of HTN was no different between controls and patients, dyslipidemia was found to be less frequent in patients with RA. In this sense, patients with RA displayed, in general, lower levels of lipid metabolism molecules. In contrast, triglycerides, lipoprotein A, and apolipoprotein B were found to be higher in patients with RA (Table 1).
As expected, the assessment of ESR and CRP revealed statistically significantly higher levels in patients with RA than in controls. Patients from our series had moderate-active disease as shown by DAS28 (3.7 ± 1.2), and 57 (38%) were taking prednisone [a median dose of 5, interquartile range (IQR) 5–6 mg/day]. Disease duration was 6.6 years (IQR 3.3–13.9), and 59% and 72%, respectively, were positive for both the anticitrullinated protein antibodies and rheumatoid factor. In addition, while 85% of the patients were taking disease-modifying antirheumatic drugs, 23% were receiving biologic therapies (13% of them anti–TNF-α agents).
Differences in carbohydrate metabolism molecules and IR indices between patients with RA and controls
HOMA2-IR indices, calculated with both insulin and C-peptide, were different between patients and controls (Table 2). HOMA2-%S was lower in patients with RA than in controls after adjusting for traditional IR-related factors and prednisone intake (105 ± 53 vs 108 ± 75, p = 0.003). Similarly, HOMA2-IR (1.65 ± 1.69 vs 1.27 ± 0.82, p = 0.011) and HOMA2-%β (134 ± 69 vs 111 ± 45, p < 0.001) were found to be higher in patients with RA in the univariate analysis. However, these differences were lost after adjusting for covariates.
When HOMA2 indices were calculated using C-peptide, the differences between patients with RA and controls were greater. In this regard, all comparisons showed higher HOMA2-IR and HOMA2-%β indices and lower HOMA2-%S in patients with RA even after multivariate analysis (Table 2).
Whereas glucose serum levels did not achieve statistically significant differences between controls and patients, insulin (13.0 ± 13.4 vs 9.8 ± 6.5 u/ml, p = 0.079) and C-peptide serum levels (3.37 ± 2.94 vs 1.53 ± 0.77 ng/ml, p < 0.001) were found to be upregulated in patients with RA. Moreover, the insulin to C-peptide ratio was lower in patients with RA, even in the multivariable analysis (0.08 ± 0.02 vs 0.14 ± 0.07, p < 0.001). These differences were still present after multivariate adjustment by glucocorticoid intake.
Propensity score–based method analysis to further study these differences disclosed similar statistically significant results.
Differences in the relationship of traditional IR factors with insulin, C-peptide, and IR indices in RA and healthy populations
Traditional IR-related factors strongly correlated with glucose, insulin, C-peptide, HOMA2-IR, and HOMA2-%β in both patients with RA and controls (Table 3). However, Pearson’s correlation coefficients were, in general, lower in patients with RA. This shows that the variability in glucose homeostasis molecules or IR indices had a weaker association with traditional IR-related factors in patients with RA. Regarding this, the relationship of waist circumference with glucose, C-peptide, and IR indices disclosed statistically significant lower correlation coefficients in patients with RA than in controls. The associations of BMI with glucose, insulin, C-peptide, and both HOMA2-%β and HOMA2-IR were also weaker in patients. Regarding lipid profiles, triglycerides (a factor strongly associated with IR) were significantly less related to glucose in patients with RA. Moreover, HDL cholesterol was negatively and less related to glucose, C-peptide, and HOMA2-%β index in patients with RA than in controls.
Interestingly, although CRP was significantly and positively associated with glucose, insulin, C-peptide, and HOMA2 indices in controls, as well as with HOMA2-%β in patients with RA, statistical analysis revealed a significant difference between patients and controls, showing a lower correlation in patients.
All these comparisons were performed with the predictive data obtained following adjustments for age, sex, waist circumference, dyslipidemia, statins, antihypertensive treatment, CRP, cholesterol levels, and glucocorticoid intake. An extended version of Table 3 is shown in Supplementary Table 1, available with the online version of this article.
Differences in the correlations of glucose, insulin, C-peptide, and IR indices between patients and controls
Linear relations between glucose homeostasis molecules (glucose, insulin, and C-peptide) and HOMA2 indices were studied in patients with RA and controls (Table 4). As expected, all the correlation coefficients between molecules and indices showed high values that reached statistical significance. However, many differences between patients and controls were observed. Except for the relation between HOMA2-IR and C-peptide serum levels, which did not reveal any differences between patients and controls, all the remaining relations between glucose homeostasis molecules and HOMA2-IR indices one to each other were found to be lower in patients with RA than in controls.
Hyperbolic relationship between insulin secretion and sensitivity in patients with RA and controls
In both patients and controls, the relationship between insulin secretion and sensitivity was hyperbolic in nature. Supplementary Table 2 (available with the online version of this article) shows that this hyperbolic relationship revealed the better r2 for controls when different nonlinear regression analyses were performed. This was not the case for patients with RA. However, nonlinear r2 and nonlinear regression analysis β coefficients showed statistically significant differences between patients and controls. In this sense, the variability explained by the curve (r2 = 0.845 vs r2 = 0.928, p = 0.001) and β coefficients (−0.74, 95% CI −0.77 to −0.70 vs −1.09, 95% CI −1.17 to −1.02, p = < 0.001) were different in patients with RA than in controls (Figure 1).
Additionally, we performed this analysis separately in patients with RA taking and not taking glucocorticoids. Patients not taking glucocorticoids compared to controls also disclosed an inferior variability explained by the curve (r2 = 0.869 vs r2 = 0.928, p = 0.001) and different β coefficients (−0.74, 95% CI −0.77 to −0.70 vs −1.15, 95% CI −1.24 to −1.06, p = 0.015; Supplementary Figure 1, available with the online version of this article).
DISCUSSION
In our present study, we show that the traditional factors associated with IR in healthy individuals are less related to IR in patients with RA compared to controls. In addition, the association between insulin secretion and sensitivity in patients with RA consists of a different nonlinear hyperbolic relationship compared to that of controls. These data may indicate that the relationship between insulin secretion and sensitivity in patients with RA may be distorted by the presence of certain RA-specific factors, most likely the presence of chronic inflammation.
Our findings on the differences in insulin, C-peptide, and HOMA2 indices between patients with RA and healthy controls are supported by several previous epidemiological studies of IR in RA7,8,9,10,11. Interestingly, Dessein, et al has previously documented that acute-phase response may explain IR in RA24 and that an increased incidence of diabetes is present in RA25. Our present study indicates that the upregulation of these molecules and indices occurs independently of traditional IR factors and glucocorticoid intake.
Although theoretical assumptions and in vitro experiments suggest equimolar secretion, the molar concentrations of serum–C-peptide are substantially greater than insulin in healthy subjects. This discrepancy apparently arises because C-peptide is more slowly catabolized than insulin. Whereas a large fraction of endogenous insulin is cleared by the liver, C-peptide, which is cleared primarily by the kidney and has a lower metabolic clearance rate than insulin, traverses the liver with essentially no extraction by hepatocytes. For this reason, the ratio of insulin to C-peptide has been assumed to reflect hepatic insulin extraction.
A number of studies have suggested that reduced hepatic extraction of insulin is a major factor in the pathogenesis underlying the hyperinsulinemia found in type 2 diabetes26 and obesity27. Although both insulin and C-peptide were higher in patients with RA, their ratio was lower in patients than in controls. We believe that the hepatic extraction of insulin may be amplified in patients with RA. Therefore, unlike other IR-related states, insulin clearance may also be greater in RA. Augmented production of C-peptide in patients with RA may also account for the lower insulin to C-peptide ratio.
Our results indicate that concomitant comorbidities associated with IR or cardiovascular risk factors have less effect on HOMA2 indices and glucose homeostasis molecules in patients with RA than in controls. Our findings support the contention that IR in RA may be mediated by disease-specific factors such as a chronic proinflammatory state. Nevertheless, this does not exclude the relevance of the traditional comorbidity factors in the development of IR. A report28 has disclosed that in established and treated RA, traditional risk factors, specifically excess adiposity, play more of a role in predicting skeletal muscle insulin sensitivity than do systemic inflammation or other disease-related factors. We think that both chronic inflammation and therapy can modify the degree and the way in which cardiovascular disease risk occurs in RA. Rather than weighing traditional risk factors and inflammation separately, we believe a more effective approach involves determining how such risk factors are modified by inflammation in patients with RA29.
Interestingly, in our work, CRP’s relationships with insulin, C-peptide, and HOMA indices were significantly stronger in controls compared to subjects with RA. This finding is in agreement with a previous report11 in which the association between CRP with IR was also weaker in the RA group. It is possible that proinflammatory cytokines, rather than CRP itself, may influence the presence of higher levels of IR in RA.
The hyperbolic relationship between insulin secretion and sensitivity was found to be weaker in patients with RA, in whom the variability and β coefficients were lower compared to controls. Although the precise mechanisms by which inflammatory factors interact to produce glucose intolerance in RA remain unclear, we believe that the loss of this hyperbolic relationship could be an expression or consequence of the failure of pancreatic β cells to compensate for insulin resistance in RA. Our finding is in agreement with the accepted pathogenesis underlying the loss of normal glucose tolerance in type 2 diabetes30. There is a dynamic relationship between IR and the compensatory increases in β-cell mass and β-cell glucose metabolism that occurs in this disease. When the compensatory process is adequate, normal glucose tolerance is maintained. However, when β-cell compensation fails, glucose levels rise, leading to either impaired glucose tolerance or overt diabetes.
Unlike our present study, previous reports described insulin secretion and insulin sensitivity as having an inverse, nonlinear, functional relationship that relied mostly on nonfasting data obtained from intravenous or oral glucose tolerance tests. Therefore, it is possible that basal concentrations may reflect only a single point in the complex glucose-insulin dose-response curve and thus may not provide a complete understanding of the ability of β-cells to respond to rising and falling glucose concentrations such as typically occur after eating31. Unfortunately, hyperglycemic clamps and graded glucose infusions are technically complex and cannot be assessed in a large series of patients. Moreover, they are limited by their nonphysiological route and pattern of glucose delivery. Another limitation of our study could be the lack of glycosylated hemoglobin (HbA1c) measurement. HbA1c, which reflects chronic blood glucose values, has been considered of value for the diagnosis of diabetes32. However, we believe that the cutoff of glycemia < 7 mmol/l used in patients and controls has reasonably excluded the presence of diabetes among the participants.
IR is present in patients with RA and cannot be completely explained by concomitant IR-related comorbidities as was observed in healthy individuals. The hyperbolic, nonlinear relationship of insulin secretion and sensitivity in subjects with RA differs from that of controls. The concept of RA as an IR-driven state warrants further research in this field.
ONLINE SUPPLEMENT
Supplementary material accompanies the online version of this article.
Footnotes
* Drs. Ferraz-Amaro and González-Gay share senior authorship and both are corresponding authors for this study.
This work was supported by grants from the Spanish Ministry of Health, Subdirección General de Evaluación y Fomento de la Investigación, Plan Estatal de Investigación Científica y Técnica y de Innovación 2013–2016 (to IFA) and by the Fondo Europeo de Desarrollo Regional [FEDER; Instituto de Salud Carlos III (ISCIII) PI14/00394, PI17/00083, PI15/00521]. Professor González-Gay’s research was supported by European Union FEDER funds and by the Fondo de Investigación Sanitaria (grants PI06/0024, PS09/00748, PI12/00060, and PI15/00525) of the Instituto de Salud Carlos III (ISCIII, Health Ministry, Spain). Other support came from the RETICS Programs RD12/0009 (RIER) and RD12/0009/0013 from the ISCIII, Health Ministry, Spain.
- Accepted for publication July 18, 2018.