Parenteral Nutrition–Associated Hyperglycemia in Non–Critically Ill Inpatients Increases the Risk of In-Hospital Mortality (Multicenter Study)

This multicenter study involved 19 hospitals in Spain (16 university hospitals and 3 nonuniversity hospitals). The study included all hospitalized noncritically ill patients (i.e., patients in the nonintensive care unit [ICU] setting) who started TPN as a sole source of nutrition between September and December 2010. Patients were excluded if they were in ICUs, receiving parenteral nutrition together with enteral nutrition, pregnant, or <14 years of age. The study was approved by the Research Ethics Committee of Carlos Haya Regional University Hospital, and all the participants gave written informed consent.

The TPN formula at all the hospitals was provided as a total nutrient admixture (3 in 1) solution containing carbohydrates, proteins, and lipids. All patients receiving TPN were seen daily by a member of the hospital nutrition unit, who made adjustments in accordance with the relevant guidelines (14,15). Prospective measurements were made of capillary blood glucose levels every 6 h, but if the blood glucose levels were <140 mg/dL, the measurements were made every 8 h. The blood glucose monitor used was the usual model in each hospital. If a patient had hyperglycemia, insulin treatment was started, following consensus recommendations (12).

Data were recorded on demographic variables; diagnosis on admission; prior comorbidity (history of kidney or liver failure, respiratory or cardiac disease, transplantation); anthropometric data (weight, height, BMI); type of TPN; concomitant prescription of steroids, somatostatin, tacrolimus, or cyclosporin; and nutritional assessment by subjective global assessment (SGA) before starting TPN (16). The mean daily insulin dose was also recorded.

Before starting the TPN infusion, a blood sample was drawn to measure the glycated hemoglobin, following the international recommendations for standardization of the HbA_1c measurement (17). Measurements were also made of fasting plasma blood glucose, albumin, and C-reactive protein (CRP) levels (with an autoanalyzer) at the laboratories of each hospital.

Patients were considered to have known diabetes if they had a documented history of diabetes. Patients were considered to have unknown diabetes if there was no record of having had diabetes but the HbA_1c was ≥6.5% (18). Any other patient with fasting plasma glucose levels ≥126 mg/dL was considered to have hyperglycemia without diabetes. Hypoglycemia was considered to be a capillary blood glucose level <70 mg/dL. The blood glucose variance was estimated from the SD of the mean blood glucose levels and their coefficient of variation.

The primary end point was all-cause in-hospital death. Other clinical outcome parameters were the number of days in the hospital and infectious complications, as recorded on the patient charts.

The comparisons between the qualitative variables were done with the χ² test, with Fisher correction when necessary. The distribution of the quantitative variables was examined with the Kolmogorov-Smirnov test. The differences between the quantitative variables were analyzed with the Student t test or ANOVA for two or more samples, respectively. Nonparametric tests (Mann-Whitney or Kruskal-Wallis) were used when the study variables did not follow a normal distribution. The significant associations were later included in multivariate logistic regression models, controlling also for other variables such as age, sex, the presence of diabetes or hyperglycemia before the administration of TPN, insulin dose per kilogram body weight, and glycated hemoglobin level. For all the calculations, significance was set at P < 0.05 for two tails.

The study included 605 patients, with a mean of 32 patients per hospital (range 19–55). The characteristics of the patients and the TPN infused are shown in Table 1. The mean HbA_1c according to diabetes status was 5.3 ± 0.5% normal (n = 308); 5.7 ± 0.4% hyperglycemia without diabetes (n = 166); 7.2 ± 1.0% unknown diabetes (n = 23); and 6.6 ± 1.1% known diabetes (n = 108). TPN was started 10 ± 14 days after admission; 433 (71%) patients received insulin at some time during the TPN infusion (55% subcutaneously, 36% in the bag and subcutaneously, and 9% with insulin perfusion independently of TPN).

There were 58 in-hospital deaths (9.6%). Table 2 shows the characteristics of the patients who died and of those who survived. Significant differences were found in age, length of stay, BMI, total grams of carbohydrates infused, mean capillary blood glucose level, SD of the blood glucose levels, CRP and albumin levels, degree of malnutrition, steroid therapy, admission diagnosis, prior comorbidities, and the development of infectious complications during admission. However, the presence of diabetes or hyperglycemia before starting TPN was not associated with a greater mortality (Table 2) or a longer hospital stay. Furthermore, no differences were found for death or mean hospital stay between the patients with diabetes and those with hyperglycemia without diabetes. Of the 605 patients, 41 (6.9%) had blood glucose levels <70 mg/dL at some time; 57 (9.4%) received octreotide or somatostatin, and 13 (2.1%) received tacrolimus or cyclosporin. However, no association was found between the presence of hypoglycemia or treatment with octreotide, somatostatin, or immunosuppressive drugs and a greater risk of mortality. Furthermore, no significant differences were found in mortality according to the type of amino acids or lipids infused.

Table 3 summarizes the characteristics of the patients according to their mean capillary blood glucose levels (<140 mg/dL, 140–180 mg/dL, and >180 mg/dL) on all the days TPN was infused. Significant differences were found in age, BMI, total kilocalories given in the TPN, mean capillary blood glucose level, HbA_1c, SD of the capillary blood glucose levels, coefficient of variation of the blood glucose levels, CRP, units of insulin given per kilogram body weight, and type of insulin treatment used.

Table 4 shows the logistic regression data for the risk of in-hospital death. Age, CRP, SGA, mean blood glucose levels, grams of carbohydrate infused, admission diagnosis, and infectious complications all differed significantly after adjusting for the other variables.

The adjusted multivariate analysis of the data showed that an increased serum glucose level during infusion of TPN (mean values >180 mg/dL) is a risk factor for an increased risk of in-hospital death in noncritically ill patients. These findings confirm and widen the results of other studies because they were carried out at a single center with fewer patients and included retrospectively both ICU patients and mixed (3–6,8,11) or solely noncritically ill patients (9) or patients receiving allogeneic hematopoietic stem cell transplantation (7,10).

In the past, stress hyperglycemia often was believed to be a useful adaptive response, but hyperglycemia has been associated with increased mortality and morbidity in a variety of medical conditions and patient populations (19). This risk could be raised with effect from blood glucose levels above just 108–110 mg/dL (11,20). The patients in the present study with blood glucose levels >180 mg/dL had a 5.6 times higher risk of death than those having blood glucose levels <140 mg/dL. In addition, this trend toward greater mortality was also seen in patients with blood glucose levels of 140–180 mg/dL (losing significance after controlling for other variables).

The total number of grams of carbohydrates infused was associated with the risk of in-hospital death. The presence of diabetes, however, was not significant, as has been found in other studies (6,9,11), suggesting that hyperglycemia is the factor most contributing to the risk of complications and death. The HbA_1c also did not contribute significantly in the logistic regression model.

In the present study, we defined the presence of diabetes and hyperglycemia before TPN infusion not only from the chart data, but also from the laboratory findings. This approach gives more value to the data and could, in part, explain the differences in earlier studies, finding that the presence of diabetes sometimes increased (5) or did not increase (8,9,11) the risk for mortality. Indeed, the prevalence of diabetes may be greatly underestimated if it is based solely on the clinical history or chart data (21).

Hyperglycemia (chronic and acute) and insulin resistance could favor prothrombotic and proinflammatory alterations as well as increase oxidative stress and impair chemotaxis, leukocyte adhesion and transmigration, and complement activation (22). Clinically, the presence of hyperglycemia during TPN infusion has been associated with an increase in infectious complications in some studies (5,6,8) but not all (3,9–11). Moreover, the use of intensive insulin therapy in noncritically ill patients seems to reduce the risk of infections (23). In the present study, the incidence of infectious complications was greater in patients with higher mean blood glucose values and in those who died, strengthening the idea of the role of these complications in outcome and mortality.

In humans, van Der Voort et al. (4) found that even low amounts of infused glucose were associated with increased mortality when glucose levels were not controlled. Part of the disparity between the results in the different papers on intensive treatment in ICU patients (19,24,25) could be the result of the nutritional therapy used in the different studies.

Hypoglycemia is the most common complication associated with inpatient insulin therapy, and it could increase mortality (26). In the present study, the prevalence of hypoglycemia was very low, and in fact, it was not associated with worse outcomes. The fact that the patients were not critically ill (less severe), that the patients started from a low baseline HbA_1c (even the diabetic patients), and the mode of delivery of the insulin (mostly in the bag or subcutaneously) may have influenced the results.

Glycemic variability has been suggested to be a significant independent predictor of ICU and hospital mortality (27,28). The SD in the present study was higher in patients who died, although the association lost significance after correcting for the other variables in the multivariate model. Although the variability is a factor that can contribute to the prognosis of these patients, hyperglycemia per se would appear to be the most contributing factor in noncritically ill patients receiving TPN.

Because the present sample comprised a heterogeneous population of noncritically ill patients from different centers, we used as markers of severity the measurement of CRP and albumin levels, which are also associated with worse outcomes and mortality in inpatients receiving artificial nutrition (29). As expected, the albumin levels were lower and the CRP levels higher in the patients who died, and CRP levels were higher in relation to blood glucose control. After adjusting for other variables, CRP level still contributed significantly in the multivariate model, strengthening the possible role of inflammation and the severity of the underlying disorder on complications and mortality in noncritically ill patients receiving TPN.

Malnutrition is associated with worse outcomes in hospitalized patients (1,30) and, therefore, could be an important confounding factor when interpreting the results of other studies that either failed to evaluate it or just included the BMI (4–6,8–11). The SGA is a simple method, and it has been used in many studies to predict adequately morbidity and mortality in hospitalized patients (16,31). In the present study, the risk of dying was three times greater in the severely malnourished patients compared with the normally nourished patients. Malnutrition together with hyperglycemia could be one of the most conditioning factors related to poor prognosis.

The use of special amino acid (e.g., enriched with glutamine) or lipid (based on olive oil or supplemented with omega-3 fatty acids) formulas may be beneficial to prevent or treat hyperglycemia (32,33), and the use of steroids could increase it (34), possibly increasing morbidity and mortality. We found no differences that depended on the type of macronutrient used, although the patients in the present series who received corticosteroids experienced greater mortality; nevertheless, this association was not significant in the multivariate analysis. The presence of accompanying diseases before hospital admission was a condition of in-hospital mortality, as in other studies (11).

The present study is not exempt from limitations. First, the blood samples were not centralized or the same blood glucose monitor was not used, which could contribute to small under- or overestimates of the real blood glucose values. Second, apart from infections, no other complications were recorded during the admission. Third, we did not establish a causal relation between capillary blood glucose levels on admission and mortality.

In conclusion, the results show that hyperglycemia in noncritically ill patients receiving TPN is associated with increased in-hospital mortality. The data suggest that the goal of metabolic control in noncritically ill patients (with or without diabetes) receiving TPN should be to reach a mean blood glucose level of <180 mg/dL. This study opens the door to further prospective studies in noncritically ill patients to determine whether stricter blood glucose control during TPN infusion improves the outcome for the patients and reduces mortality.

The Spanish Society of Endocrinology and Nutrition provided help with the publication expenses of the manuscript.

The funder played no role in the conduct of the study, collection of data, management of the study, analysis of data, interpretation of data, or preparation of the manuscript.

No potential conflicts of interest relevant to this article were reported.

G.O. and M.J.T. contributed to the conception and design of the study; acquisition, analysis, and interpretation of the data; statistical analysis; and drafting of the manuscript. J.Oc., C.C.-G., M.D.B.-P., A.V.-C., C.A.-I., J.Ol., M.C.C.-G., A.G.-M., F.B.-R., R.P.Q.-T., L.Ca., P.M., L.Ch., R.B., P.P., M.F., A.Z., J.P., M.D., M.J.C., A.V.-B., J.R.U., C.A.-V., A.R., I.B., P.G.-P., A.M.-G., E.M., D.O., J.L.P., and M.C.T. contributed to the data acquisition and critical review of the manuscript. G.O. and M.J.T. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Parts of this study were presented in abstract form at the 27th Congress of the Spanish Society of Parental and Enteral Nutrition, Madrid, Spain, 8–11 May 2012.

The authors thank the patients for their participation in the study.