Acute hyperglycemia alters the immune response. In vitro studies show that high glucose levels augment endothelial activation (1), leukocyte adhesion (2), and the expression of alloantigens (3). In addition, high glucose levels in the immediate period after surgery potentiate ischemic injury (4) and possibly initiate wound infection (5). The Australia and New Zealand Dialysis and Transplant registry (ANZDATA) has recently reported that renal transplantation patients with diabetes have more acute rejection (6). We hypothesize that the early exposure and response of allograft tissue to hyperglycemia could in part explain this increased risk of rejection. This pilot study examines the relationship of perioperative glucose control to acute rejection in patients with diabetes.
A retrospective review was made of the records of each of the 423 patients who underwent their first cadaveric renal transplant (CD1) at The Queen Elizabeth Hospital, Adelaide, Australia, between January 1990 and July 2000. Patients with primary failure, death without graft function, or graft loss due to vascular complications were excluded. All patients who commenced treatment with a steroid-sparing protocol, consisting of cyclosporine A and mycophenolate or azathioprine, were identified (n=243). In this protocol, cyclosporine A (8 mg/kg/d or 5 mg/kg/d with diltiazem) mycophenolate (2 g/d) or azathioprine (2 mg/kg/d) were first given orally 6 to 8 hr after transplantation. Every patient received an intravenous bolus dose of 1 g methylprednisolone before surgery and 500 mg (IV) the following morning without any continued oral steroid therapy. Only CD1 patients, who were unsensitized to panel reactive antigens (PRA) (peak<50% and negative T and B cell cross-match) were selected for this regimen. From this cohort, all established diabetic patients were studied (n=50). Diabetes mellitus was defined as the presence of chronic glucose intolerance at the time of operation. Patients who developed de novo or posttransplant diabetes mellitus (PTDM) were not designated as diabetic for the purposes of this study but are studied elsewhere (7). Perioperative glycemic management was not standardized.
All available postoperative glucose levels for the first 100 hr after surgery were obtained from patient records. Venous blood glucose levels were used for the first perioperative reading. Capillary samples were taken at 1- to 2-hr intervals over the first 48 hr and 1- to 5-hr intervals on subsequent days. The mean glucose levels were calculated from the area under the curve of glucose versus time. Poor glucose control was arbitrarily defined as a mean capillary glucose level >11.2 mmol/L (200 mg/dL), after Furnary et al. (5).
Transplantation data, including nine continuous variables (donor age; recipient age; duration of operation; warm and cold ischemic times; PRA-peak; PRA-current; and human histocompatibility leukocyte antigen (HLA) haplotype matching and mismatching) and six categorical variables (donor ethnicity and gender, recipient ethnicity and gender, immediacy of graft function, and whether the patient received mycophenolate or azathioprine), were obtained from a common database (Table 1). Posttransplantation records of all 50 patients were then examined for the presence of acute allograft rejection. Acute rejection was said to occur if biopsy-proven or clinical rejection occurred within 20 days of transplantation. Late acute rejection (after 20 days) was felt unlikely to be related to early hyperglycemia and was not specifically studied. If a biopsy was not performed, clinical rejection was defined retrospectively by an otherwise unexplained and sustained rise of more than 10% from the predicted plasma creatinine responsive to adjunctive immunosuppressive therapy.
Previous studies have shown early perioperative glycemic control to be closely related to wound infection after cardiac surgery (5). To confirm the presence of clinically significant early hyperglycemia in our patients, 11 additional historical, demographic, and surgical variables that might also be associated with postoperative infection (POI) were collected from patient records. These included seven categorical variables: type of diabetes (Type 1, Type 2– requiring insulin, Type 2–not requiring insulin), type of dialysis (hemodialysis or continuous ambulatory peritoneal dialysis), smoking history, nasal Staphylococcus aureus carriage, historical comorbidity (hypertension and congestive heart failure), and the presence of a fluid collection within the tranplant wound on the first postoperative day) and five continuous variables: recipient body mass index, admission leukocyte count, admission albumin, estimated blood loss, and amount of postoperative weight gain. All patients received standard perioperative antibiotic prophylaxis with intravenous cephalothin or vancomycin. Postoperative infection was defined by the synchronous occurrence of a positive culture of bacterial pathogens and symptoms of infection together with a fever >37.5°C or the use of antibiotics (not as prophylaxis) within 20 days of transplantation.
Each of these variables, considered likely to influence POI and rejection, was then entered into a multivariate logistic regression. Because of the relatively small data set (n=50), predictor variables were defined by backward selection using the Akaike information criterion. The functional form (in particular, nonlinearity) of continuous variables in the final model was checked graphically, using partial residual plots, and formally, by parametric and nonparametric means. Categorical (cutpoint) analysis of continuous variables was not primarily used except when testing for first-order interactions. Logistic model performance was assessed by indices of calibration (Hosmer Lemeshow C ‘deciles of risk’ statistic) and discrimination (area under receiver operating characteristic (ROC) curve). Confidence intervals (95%) of the C statistic and ROC curve area were computed using the bootstrap method (BCa, bias corrected and accelerated). The parameters in the final model were estimated as risk ratios (generalized linear model with Bernoulli family and log link), to correct the odds ratio interpretation under conditions of common prevalence (>10%) in the study population. Univariate results are expressed as ±95% confidence interval. Stata® statistical software, version 6.0 (1999) was used.
Mean age was 52.0±2.7 years, and 50% of patients were men. Fifty-eight percent of diabetic patients were aboriginal, including 16 men and 18 women. Patients with Type I diabetes comprised 18% of the group. Eighty-two percent had Type II diabetes controlled by insulin (8%) or by oral agents and/or diet (92%). Eighty-four percent had renal failure attributed to diabetes. Patients had a mean of 35±5.8 capillary glucose samples in the first 100 hr following surgery (range=26–78, median=41).
POI occurred in 35 of the 50 diabetic patients (70%) at a mean of 10.8±2.3 days after transplantation. Of the variables assessed, only glycemic control, duration of operation, and the presence of preoperative leucocytosis were significantly associated with POI on multivariate analysis. Using the mean glucose level over the first 100 hr after surgery as the best measure of postoperative glycemic control, a predictive model was generated (Hosmer-Lemeshow C=6.627, P =0.577, ROC=0.82±0.5). Notably, body mass index, ethnicity, and type of dialysis did not influence the occurrence of POI in this study. The mean glucose levels over the first day after surgery were 9.3±0.7 in patients without POI and 13.8±1.2 mmol/L in patients with POI (P <0.001). Glucose levels over the first 48 hr were 9.6±0.7 mmol/L in patients without POI and 13.3±1.2 mmol/L in patients with POI (P <0.001). Although an increased incidence of posttransplant infections in patients with diabetes has been previously described, this is the first time that glycemic control per se has been directly linked with infection following renal transplantation. Every patient with poor glycemic control in this study (mean blood glucose level over the 100 hr following surgery >11.2 mmol/L, n=23) experienced postoperative infection. Systemic signs of infection occurred significantly later than the hyperglycemia in the immediate postoperative period. Although it is possible, it seems unlikely that occult infection would manifest as hyperglycemia alone 6 to 10 days before clinical presentation. Our findings are consistent with other forms of major surgery, showing that early glycemic control has a causal relationship with postoperative infection (5).
Acute rejection occurred in 20 of the 50 patients with diabetes (40% overall) at a mean of 7.7±2.6 days after transplantation. Eighteen patients (92%) had biopsy-proven rejection and two had clinical rejection. The final predictive model used only three continuous variables; age, operative time, and mean glucose levels after surgery were independent predictors for rejection (Table 2). Although more patients with rejection used azathioprine than used mycophenolate, this did not reach significance independent of other variables. No evidence of nonlinearity was demonstrable using either parametric (fractional polynomials) or nonparametric (generalized additive model [GAM] with cubic smoothing spline) modeling, and no interactions were evident. The C statistic was 3.07 (95% CI; 1.55–3.63) with corresponding P-values 0.93 (0.89–0.99) and ROC curve area 0.86 (0.77–0.94). As with POI, a statistically significant difference in glycemic control was present from the moment patients returned from the operating room and continued over all time periods (Fig. 1). The mean glucose levels over the first day after surgery were 10.8±0.8 mmol/L in patients without rejection and 15.0±1.8 mmol/L in patients with rejection. Mean glucose levels during the first 48 hr were 10.6±0.7 mmol/L in patients without rejection and 14.4±1.9 mmol/L in patients with rejection. Only 3 of the 27 patients with optimal glycemic control in the 100 hr after surgery (mean <11.2 mmol/L) had rejection episodes, compared with 58% of diabetic patients with poor control (mean >11.2 mmol/L).
Although no previous study has specifically addressed the influence of early hyperglycemia on the incidence of allograft rejection, there is an increased incidence of acute rejection in patients with diabetes (6). Our data suggest that it is the hyperglycemic subgroup that may have the highest rejection rates, raising the possibility of a causal link between glycemic control and allograft rejection. Although our numbers are small, the independent correlation is particularly striking. In this study, hyperglycemia was a stronger predictor of rejection than traditional risk factors.
Acute rejection is thought to be initiated in the early postoperative period by antigen presentation, possibly in response to allograft inflammation and injury. Nonspecific factors such as graft ischemia have been linked to acute rejection (8). We have recently suggested that hyperglycemia may also increase the risk of allograft rejection by one of the following three broad mechanisms (7). First, high glucose levels may exacerbate ischemic injury and apoptosis. Second, antigen presentation is increased in hyperglycemia. Third, hyperglycemia induces an exaggerated inflammatory response to ischemia/reperfusion and rejection. We found no association between POI and rejection in our diabetic population, independent of glycemic control. POI occurred significantly later than rejection episodes (10.8 vs. 7.7 days, P <0.05), although this does not exclude a link between preclinical infection, rejection, and (because optimization of perioperative glucose levels reduces POI) glycemic control.
Patients with rejection had higher mean glucose levels at every time point following surgery, from the moment they left the operating room to 100 hr after transplantation (Fig. 1). Because of this lack of any conversion, it is possible that poor glycemic control may only be a marker for another difference between patients who do and do not reject. In particular, recalcitrant hyperglycemia following steroid use and surgery suggests the presence of the insulin resistance syndrome. Hypertension, dyslipidemia, hyperinsulinism, and increased levels of circulating advanced glycation end-products, leptin, and proinflammatory cytokines characterize this metabolic milieu (to which graft tissue would be newly exposed). These may act, by themselves or in combination with hyperglycemia, to up-regulate allograft injury or rejection. Patients with the insulin resistance syndrome also possess abnormalities of the innate immune system, including an augmented cytokine responsiveness (9) that may predispose to rejection. A recent study has shown that pretransplant serum C-reactive protein, a recognized marker of inflammatory responsiveness independently predicts allograft rejection (10). Certainly, patients with the worst glycemic control in our study received more insulin, implying that they also had greater insulin resistance. In these patients, large doses of insulin are often administered in the attempt to control glucose levels, resulting in significant hyperinsulinism, alone capable of promoting monocyte activity and endothelial activation and growth and raising blood pressure (11). However, patients who achieved optimal glycemic control had significantly less rejection, despite receiving large insulin doses. This suggests that hyperglycemia is more than just an epiphenomenon. Nevertheless, interventions that improve both postoperative insulin resistance and glycemic control (such as optimizing sugars weeks before transplantation and minimizing the use of high-dose steroids and dextrose at induction) may be particularly valuable in transplant patients with diabetes.
This pilot study has been a comparatively small ex post facto survey. Serial hemoglobin A1c measures using identical methodology were not available for patients in this study. Postoperative therapy for hyperglycemia was not standardized, although attempts were made to optimize postoperative glucose levels in every case. Some patients received insulin infusions and others oral hypoglycemic therapy. Patients with higher glucose levels were more likely to receive more insulin. In addition, monitoring of either cyclosporine A or mycophenolate did not form part of this study. Certainly, pharmacokinetics may be altered in patients with diabetes, although it seems unlikely that therapeutic drug levels would be confounded by glucose control.
Although hyperglycemia is not the only risk factor for rejection, it appears to be one susceptible to intervention. Intensive perioperative glycemic control is now practical and safe and has already demonstrated cost benefits in preventing wound infections (5). Our pilot study has shown for the first time that early hyperglycemia in patients with diabetes may be associated with an increased risk of allograft injury. We believe prevention of hyperglycemia and attention to insulin resistance may reduce postoperative infection and decrease rejection episodes. Because diabetes is the most common reason for transplantation in the western world, a randomized clinical trial of intensive glycemic control following transplantation would seem overdue.
1. Takami S, Yamashita S, Kihara S, et al. High concentration of glucose induces expression of intercellular adhesion molecules-1 in human umbilical vein endothelial cells. Atherosclerosis 1998; 138: 35.
2. Morigi M, Angioletti S, Imberti B, et al. Leucocyte-endothelial interaction is augmented by high glucose in an NF-kB-dependent fashion. J Clin Invest 1998; 101: 1905.
3. Pavlovic D, van de Winkel M, van der Auwera B, et al. Effect of interferon-gamma and glucose on major histocompatibility complex class I and class II expression by pancreatic beta- and non-beta-cells. J Clin Endocrinol Metab 1997; 82: 2329.
4. Podrazik RM, Natale JE, Zelenock GB, D’Alecy LG. Hyperglycemia exacerbates and insulin fails to protect in acute renal ischemia in the rat. J Surg Res 1989; 46: 572.
5. Furnary AP, Zerr KJ Grunkemeier GJ, Starr A. Continuous intravenous insulin infusion reduces the incidence of deep sternal wound infection in diabetic patients following cardiac surgery. Ann Thorac Surg 1999; 67: 352, discussion 360.
6. Disney APS, ed. Twenty-second report of the Australia and New Zealand dialysis and transplant registry (ANZDATA). Adelaide, South Australia: , 1999.
7. Thomas MC, Mathew TH, Russ GR, Rao MM, Moran J. Early peri-operative hyperglycemic and renal allograft rejection in patients without diabetes. BMC Nephrol 2000; 1: 1.
8. Lu CY, Penfield JG, Kielar ML, Vazquez MA, Jeyarajah DR. Hypothesis: is renal allograft rejection initiated by the response to injury sustained during the transplant process? Kidney Int 1999: 55: 2157.
9. Fernandez-Real JM, Ricart W. Insulin resistance and inflammation in an evolutionary perspective: the contribution of cytokine genotype/phenotype to thriftiness. Diabetologia 1999; 42: 1367.
10. Perez RV, Brown DJ, Katznelson SA, et al. Pre-transplant systemic inflammation and acute rejection after renal transplantation. Transplantation 2000; 69: 869.
11. Solini A, DeFronzo RA. Hypertension, cardiovascular disease, diabetes mellitus and diabetic nephropathy: role of insulin resistance. In: Mogensen CE, ed. The kidney and hypertension in diabetes mellitus, 3rd ed. London: Kluwer Academic Publishers, 1996: 61.