Diabetes does not affect outcome in patients with Enterobacteriaceae bacteremia

Background There is limited information about the effect of diabetes on the prognosis of patients with bacterial infections. We performed a retrospective cohort study to investigate possible correlations between diabetes and prognosis in patients with Enterobacteriaceae bacteremia. Methods We reviewed the medical charts of 1112 patients who were treated at a community teaching hospital for Enterobacteriaceae bacteremia from January 1997 through June 2007. Factors associated with in-hospital mortality were analyzed by logistic regression analysis. Results Among the 1112 patients with Enterobacteriaceae bacteremia, 181 (16.3%) were diabetic patients; 90 patients (8.1%) died while in the hospital. Compared to non-diabetic patients, diabetic patients were older (75.4 ± 11.9 years vs. 70 ± 16.6 years, p < 0.001) and had more comorbidities. However, mortality among diabetic and non-diabetic patients was not different [7.2% vs. 8.2%, RR 1.13; 95% CI (0.67–1.9); p = 0.39]. In a multivariate analysis, the variables associated with in-hospital mortality were age, the origin of the bacteremia, and the presence of immunosuppression. Diabetes was not associated with outcome. Conclusion In this cohort of patients with Enterobacteriaceae bacteremia, diabetes was not associated with a poorer prognosis.


Background
Enterobacteriaceae are the dominant causal agents of Gram-negative bacteremia and are a major cause of morbidity and mortality [1]. The prevalence of diabetes is increasing worldwide [2]; considering the predisposition of diabetes patients to Enterobacteriaceae bacteremia [3], the number of patients with both diabetes and bacteremia caused by Enterobacteriaceae is likely to increase as well. However, information about the effect of diabetes on the prognosis of patients with infections in general, and with Enterobacteriaceae bacteremia in particular, is limited. In addition, some data are contradictory [3][4][5][6]. It is significant to determine whether diabetes has a negative impact on the prognosis of Enterobacteriaceae bacteremia so this would support the convenience of a more aggressive approach for these patients. We performed a retrospective cohort study to identify possible correlations between diabetes and prognosis in patients with Enterobacteriaceae bacteremia.

Methods
The Sierrallana Hospital is an adult acute-care center in Torrelavega, a city in the province of Cantabria, Spain that forms part of a health district of 160,000 inhabitants. It is a teaching institution with 250 beds that has approximately 8000 admissions and 65000 assistances at the emergency service annually. It includes most major departments and specialties, except transplantation, thoracic, cardiovascular surgery and neurosurgery units. Using the microbiology laboratory database, we identified patients with Enterobacteriaceae bacteremia who were treated at Hospital Sierrallana, from January 1997 through June 2007. Only the first episode of bacteremia in a particular patient, accounted at our hospital during the period of study, was included. Enterobacteriaceae strains recovered from blood samples were identified and tested for antimicrobial susceptibility by the automated testing systems MicroScan (Dade Behring, Sacramento, CA, USA) or BD Phoenix (Becton-Dickinson Biosciences, Sparks, MD, USA).
A standardized data collection form was used to review the hospital records of the included patients. Diabetes was considered present if there was a previous diagnosis of type 1 or type 2 diabetes, or if diabetes was diagnosed during the patient's treatment for bacteremia. For the diagnosis of diabetes the American Diabetes Association diagnostic criteria were used [7]. Renal insufficiency was indicated by a creatinine value ≥ 2.0 mg/dL. Neutropenia was defined as an absolute neutrophil count of <500 cells/ mm 3 at the bacteremia onset. Immunosuppression was defined as the presence of neutropenia, infection with the human immunodeficiency virus or current treatment with immunosuppressive agents. Sepsis, severe sepsis and septic shock were defined according to the guidelines of Bone et al [8]. The initial empirical antimicrobial therapy was considered appropriate if the antibiotics were administered within 24 h of acquisition of a blood culture sample and included at least one antibiotic that was active in vitro against the causative microorganism(s). Therapy with uri-nary antiseptics such as norfloxacin, fosfomycin, pipemidic acid or nalidixic acid was considered inadequate.
A significance level of 0.05 (2-sided) was used for all tests. Relative risk (RR) was calculated as the ratio of the incidence rate in patients exposed to a risk factor to the incidence rate in those not exposed. Univariate analyses and logistic regression were performed to identify factors associated with patient death. Factors were added to the model using variables with a p-value <0.1 in the univariate analysis. Diabetes was included as a variable in the regression model independent of the p-value. The statistical analyses were performed using SPSS software, version 14.0. The study was approved by the Comité Etico de Investigación Clínica de Cantabria.

Results
During the study period, 1134 cases of Enterobacteriaceae bacteremia were identified; of these, full medical records were available for 1112 (98% Variables significantly associated with mortality in the univariate analysis are shown in Table 2. As nosocomial and community acquired bacteremia are quite different entities in terms of risk factors and prognosis a stratified analysis was performed analyzing variables associated with prognosis in these subgroups (Table 3 and Table 4). Length of hospital stay, days (mean ± SD) 13.3 ± 12 13.9 ± 15 0.64 Data are n (%) otherwise indicated *RR: Relative risk **COPD: chronic obstructive pulmonary disease    Variables that were associated with in-hospital mortality in the multivariate analysis in the whole cohort, in nosocomial cases and in community acquired cases are reflected in Table 5. The severity of sepsis was not included in the initial models [9]. Diabetes was not associated with outcome in these analyses. In a second analysis including the severity of sepsis in the whole cohort, in the nosocomial cases and in community acquired cases, diabetes was not associated with outcome.

Discussion
Diabetes has traditionally been associated with a worse prognosis for patients with infectious diseases such as Enterobacteriaceae bacteremia [3], liver abscess [4] and tuberculosis [6]. However, the prognosis of diabetic patients has improved in recent years for some procedures, such as coronary bypass grafting [10], and for some conditions, such as myocardial infarction [11]. This improvement has been attributed, at least in part, to better glucose control in diabetics. A similar improvement would be expected for the prognosis of diabetic patients with bacterial infections, which is what we found in this study. Another recent study also found no differences in the outcomes of diabetic patients with communityacquired bacteremia [5]. Improvements in glucose control in diabetic patients may account for the discordance between our results and those of Thomsen et al [3]; this group found a slightly higher risk of late death in diabetic patients in a population study of patients with Enterobacteriaceae bacteremia conducted from 1992 through 2001.
Other explanations could be related to differences in the population analyzed, comorbidities and the adequacy of antibiotic empirical treatment. We included some variables that were not analyzed by Thomsen et al, which could also contribute to the difference in results. In addition, we only evaluated the effect of diabetes on in-hospital mortality, not on 30-and 90-day mortality [3]. Our study is the first one which analyzes the mortality of diabetic patients with bacteremia performed in Mediterranean country. The protective effect of the Mediterranean diet over our diabetic patients with bacteremia cannot be discarded.
The mortality in our series is lower that previous similar studies [3]. Several factors, such as a lower incidence of comorbidity, the low proportion of patients with septic shock and the high proportion of adequate empirical treatment among our patients with Enterobacteriaceae bacteraemia may contribute to the low mortality [12]. The incidence of Enterobacteriaceae bacteremia in our series in much higher than in others [3,13]; this suggest that the lower mortality are due to a more vigilant attitude in our centre that lead to the detection of more mild cases of bacteremia.
Our study has several potential limitations. The differences in mortality could be lower that the sample size would permit to detect. As the data, including the diagnosis of diabetes has been obtained retrospectively, this could lead to information bias between exposure and out-   come, so the diagnosis of diabetes may actually have been missed in patients dying precipitously, and conversely there is a risk that a diagnosis of diabetes has been wrongly inferred from hyperglycemia in some surviving patients. We did not routinely collect measures of longterm glucose control, such as HbA1c, which would have allowed us to distinguish the effects of long-term poorlycontrolled diabetes on patient prognosis. Finally, our population was mainly elderly people with type II diabetes, thus, these results may not be applicable to patients diagnosed with type I diabetes.
Our data suggest that the presence of diabetes is not related to in-hospital mortality in patients with Enterobacteriaceae bacteremia. Improved management of diabetic patients with acute illnesses in recent years may account for the absence of an effect of diabetes on outcomes in these patients.

Conclusion
In conclusion, in our cohort of patients with Enterobacteriaceae bacteremia, diabetes was not associated with a poorer prognosis. Improved management of diabetic patients with acute illnesses in recent years may account for the absence of an effect of diabetes on outcomes in these patients.