Post-operative stress hyperglycemia is a predictor of mortality in liver transplantation

Orthotopic liver transplantation (OLT) is the established treatment for end-stage liver disease and acute fulminant hepatic failure, and more than 100,000 OLTs have been performed in Europe [1]. Advances in both medical management and surgical techniques have led to an increase in the number of long-term survivors [2].

Alterations in glucose metabolism are common among patients undergoing surgery, and are associated with increased risk of mortality and morbidity [3]. These abnormalities, particularly hyperglycaemia, are also common in critically ill patients, even those without a diagnosis of diabetes. Hyperglycaemia associated with critical illness is commonly regarded as an extreme form of “stress hyperglycaemia”, typically attributed to insulin resistance caused by endogenous and exogenous catecholamines and glucocorticoids. High levels of circulating free fatty acids inhibit peripheral glucose uptake and use, causing hepatic steatosis, which impairs liver glucose regulation in the critically ill [4, 5]. In addition, blood glucose concentrations vary markedly in critically ill patients, even when using continuous feeding and insulin infusions, and blood glucose values and incidence of hyperglycaemia have a circadian rhythm in critically ill patients [6]. Indeed, blood glucose variability can be quite different in the presence of the same mean blood glucose value.

Several researchers have consistently reported a significant association between increased glycaemic variability and worse outcome in critically ill patients [7‐12]. In their analyses, blood glucose variability is measured by using standard deviation, percentile values and successive changes in blood glucose, and by calculating the coefficient of variation. It is, however, recognised that, compared to the use of only single-moment biomarker values, serial biomarker evaluations may carry important additional information as regards prognosis of the disease under study.

We hypothesised that postoperative blood glucose concentrations would independently predict mortality in patients undergoing OLT. During patients’ stay in hospital, clinicians use longitudinal glucose measurements to gain a better understanding of disease dynamics. Accordingly, this study sought to investigate the ability of postoperative glucose profiles measured for 7 days to predict mortality in patients who underwent OLT, differentiating between those with and those without pre-existing diabetes mellitus. The relationship between glucose profiles and risk of death was modelled using flexible joint modelling of longitudinal data and time-to-event analysis.

We conducted an observational study of 642 patients undergoing liver transplantation at a single, tertiary care transplant hospital from July 1994 to July 2011. Patients were followed up until 12 July 2012. The following exclusion criteria were applied: any patient who had a previous organ transplant, had any other invasive surgery at the date of transplantation, had incomplete medical records or had died within the first 72 h. Finally, a total of 632 patients were available for study. Data were drawn from hospital medical records and transplant database. Pre-transplant variables were recorded, including age, gender, body mass index (BMI), indication for OLT, Model for End-stage Liver Disease (MELD) score, haemoglobin, haematocrit, platelet count, prothrombin time, serum total bilirubin levels, serum creatinine levels, fasting blood glucose and prior diagnosis of diabetes. In addition, the following surgical variables were studied: surgical technique; transfusion requirements; operating time; and cold ischaemia time (CIT). The postoperative period was defined as the 7 days immediately following the date of transplant. Data relating to immunosuppressive drug regimen, creatinine, insulin, parenteral nutrition (PN), continuous veno-venous haemodiafiltration (CVVHDF), use of insulin drip, and glucose were measured across the postoperative period.

This study conformed to the principles of the Helsinky Declaration, and was reviewed and approval by the Local Research Committee (code # CEIC2011/328). Written inform consent was obtained from all participants.

In this survival study among OLT patients, the main findings were: (1) plasma glucose profiles during the first week post-OLT are different in patients with and without diabetes mellitus prior to transplantation; (2) while glucose levels are associated with a higher risk of mortality among patients without diabetes, after adjusting for potential confounding variables such as age, cause of transplantation, MELD, BMI, nutrition, and haemodiafiltration, this does not occur among patients with diabetes; and, (3) the glucose profiles analysis provides better information than a single measure of glucose variability to ascertain the relationship between glucose levels and mortality.

The fact that glucose values may give rise to an increase in long-term mortality among critical patients has been previously described. The association between glucose and mortality has been the subject of study by a number of authors [16‐20], though not all report uniform results. Some authors observe that in many diseases, critical patients having higher glucose levels register higher mortality [21‐24], something that has led in turn to the implementation of different protocols aimed at achieving a strict control of glycaemia, so as to maintain glucose values within the range of normality. Such strict control does not necessarily entail a decrease in mortality, however. Indeed, in some patients this has even been observed to lead to an increase in mortality, possibly due to a higher rate of hypoglycaemia [25‐28].

With respect to glucose levels, there are also reports of different mortality patterns observed in patients with and without diabetes. Patients with diabetes tolerate higher glycaemic values, without repercussions on mortality, conceivably possessing some, as yet unknown, protective factors against stress hyperglycaemia. These facts, together with the finding that in the first week post-OLT diabetic patients display glucose profiles different to those of non-diabetic patients, led us to analyse these two groups of patients separately. As can be seen from Fig. 1, patients with diabetes started with higher plasma glucose values on the first day, which then gradually declined after the first 48 h. In contrast, non-diabetic patients started from appreciably lower figures which gradually rose until reaching a peak at 48–72 h, and then subsequently declined (all within the context of insulin being administered via continuous perfusion).

This agrees with the results of previous studies in which the presence of diabetes has been reported to afford a protective effect against hyperglycaemia [29‐31]; hence, despite the fact that patients with diabetes display higher glycaemic values, this would not affect mortality. This might perhaps be attributable to the existence of a chronic state of hyperglycaemia in such patients, which would confer a greater tolerance to the condition and so attenuate its harmful effects in situations of stress. Already in Egi’s study, it was observed that, even within the group of patients with diabetes, those who presented with higher glycated haemoglobin figures, with worse control of their disease, and had thus chronically maintained higher glucose figures, would tolerate hyperglycaemia better [32].

Despite the many studies conducted on glycaemia in a critical care setting, very few references could be found in the field of liver transplantation [33‐35], and none of them stratified patients into those with and those without diabetes. Furthermore, when glycaemic levels were studied, consideration was given only to isolated glycaemic values or summary values of variability, such as standard deviation [36] or the coefficient of variation [29], with no follow-up of glycaemia across time having been previously conducted as in our study. Moreover, joint-modelling analysis allows for a better approximation of the survival of such patients, by taking trends in glycaemic levels in the postoperative period into account.

Similarly, patients with diabetes display higher absolute glycaemic values than do patients without diabetes during the first 7 days after transplantation. Although glycaemia does not attain such high values in patients without diabetes, glycaemic levels nevertheless display a marked rise at day 3, which is absent among diabetics and could reflect glycaemic response to stress. Despite insulin being administered via continuous perfusion to maintain glycaemia figures between 120 and 180 mg/dL, the distribution adopted by glycaemia remained unchanged. This coupled with the fact that glycaemic profiles were associated with mortality solely in patients without diabetes, led us to think that the presence of this hyperglycaemic peak might possibly flag patients with the highest stress—and thus the most severe patients—and not glycaemic levels per se. This may be compatible with findings of previous studies in which correction of hyperglycaemia was observed to lead to an increase in hypoglycaemia and mortality rates. It is for this reason that, on evaluating glycaemia in post-OLT patients, we feel that account should be taken, not only of the glycaemic values but also of their trend over time, distinguishing between patients with and those without diabetes.

Glucose profiles during the first week after liver transplantation are associated with long-term mortality in patients without diabetes, after adjusting for potential confounding variables such as age, cause of transplantation, MELD, nutrition, immunosuppressive drugs, and units of insulin administered. Furthermore, these profiles are different in patients with and without diabetes.