Roghmann M. Predicting Methicillin Resistance and the Effect of Inadequate Empiric Therapy on Survival in Patients With Staphylococcus aureus Bacteremia. Arch Intern Med. 2000;160(7):1001-1004. doi:10.1001/archinte.160.7.1001
Copyright 2000 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2000
The restriction of vancomycin hydrochloride use is recommended as a measure to decrease the emergence of vancomycin resistance in gram-positive organisms; however, vancomycin also is the treatment of choice for methicillin-resistant Staphylococcus aureus (MRSA) infections. If vancomycin use is restricted to patients with documented infections due to methicillin-resistant organisms, then patients with MRSA infections may not initially receive vancomycin. This study was performed to determine factors that predict MRSA bacteremia and if ineffective empiric antibiotic therapy increased the risk of death in patients with S aureus bacteremia.
We conducted a retrospective cohort study of all patients with clinically significant S aureus bacteremia (132 episodes in 128 patients) diagnosed between October 1, 1995, and January 1, 1998, at an urban acute care Veterans Affairs medical center (approximately 200 acute care beds) in Baltimore, Md. During the study period, vancomycin was a restricted antibiotic. Empiric use had to be approved by an attending physician specializing in infectious diseases.
Compared with patients who had methicillin-sensitive S aureus bacteremia, patients with MRSA bacteremia were significantly older (70 vs 58 years; P<.01), more likely to have a history of MRSA (47% vs 6%; P<.01) and a nosocomial infection (76% vs 50%; P<.01), and less likely to use injection drugs (8% vs 32%; P<.01). In addition, compared with patients who had methicillin-sensitive S aureus bacteremia, patients with MRSA bacteremia were significantly less likely (45% vs 98%; P<.01) to receive effective antibiotic therapy during the first 48 hours of hospitalization. However, the risk of death due to ineffective empiric therapy was less than 1 (relative risk, 0.82; 95% confidence interval, 0.36-1.88) and did not change significantly when adjusted for age, occurrence of sepsis, or nosocomial infection.
The results of this study support the safety of the restriction of vancomycin use in patients with clinically significant S aureus bacteremia. However, patients with a history of MRSA are more likely to have future MRSA infections and should receive empiric therapy using vancomycin for possible S aureus infections, particularly for nosocomial infections.
THE RESTRICTION of vancomycin hydrochloride use is recommended as a measure to decrease the emergence of vancomycin resistance in gram-positive organisms,1 including enterococcus and Staphylococcus aureus; however, vancomycin also is the treatment of choice for methicillin-resistant S aureus (MRSA) infections.2 If vancomycin use is restricted to patients with documented infections due to methicillin-resistant organisms, then patients with MRSA infections may not receive vancomycin for the first 48 hours of therapy until the cultures are reported as positive for S aureus. This study was performed to determine factors that predict MRSA bacteremia in patients with S aureus bacteremia and if ineffective empiric antibiotic therapy increased the risk of death in patients with S aureus bacteremia.
This study is a retrospective cohort study of all patients with clinically significant S aureus bacteremia (132 episodes in 128 patients) diagnosed between October 1, 1995, and January 1, 1998, at an urban acute care Veterans Affairs medical center in Baltimore, Md, with approximately 200 acute care beds. Only the first bacteremic episode for each patient was included in the analysis. During the study period, vancomycin was a restricted antibiotic. Empiric use of antibiotics had to be approved by an attending physician specializing in infectious diseases. Bacteremias were significant if one or more blood cultures showed growth of S aureus (hereinafter "positive") and if the patient exhibited one or more signs or symptoms of infection.
Patient medical charts and electronic medical records were reviewed for basic demographic information, underlying diseases, prognosis of underlying diseases, characteristics of infection, treatment with an antibiotic, and clinical outcomes. The prognosis of an underlying disease was classified by the criteria of McCabe and Jackson.3 Bacteremias were classified as nosocomial if they occurred after 3 or more days of hospitalization in any health care facility or if they were due to a central vascular catheter. The patient response to bacteremia was categorized as severe sepsis if the patient met the criteria for severe sepsis specified by the American College of Chest Physicians/Society for Critical Care Medicine Consensus Conference Committee.4 Antibiotic treatment was administered during the first 5 days after the first positive blood culture was noted in the charts. Empiric antibiotic therapy was considered ineffective if the organism was not sensitive to any antibiotic given during the first 48 hours.
Categorical variables were compared by Fisher exact test or Pearson χ2, as appropriate, and continuous variables were compared by the Wilcoxon rank sum test. Two-tailed tests were performed, and P=.05 was considered statistically significant. A stratified analysis was performed using severity of infection and other variables associated with decreased survival to assess for confounding (a change in the relative risk [RR] of greater than 20%) and interaction (a Breslow-Day test for homogeneity, >0.10). Logistic regression models were used to test the association between methicillin resistance (dependent variable) and variables associated with methicillin resistance in univariate analysis (with P<.15).
The patients (n=128) had a median age of 64 years (range, 33-92 years) and were predominantly men (99%). Diabetes mellitus (36%), alcoholism (19%), and injection drug use (25%) were the most common underlying conditions. Of the 128 patients, 23 (18%) had previously had MRSA identified in clinical cultures (history of MRSA infection or colonization); 64 (50%) had a rapidly or ultimately fatal underlying disease; 83 (65%) had a prior admission to the hospital in the past year; 76 (58%) of the 132 bacteremias were caused by nosocomial infections; 40 (31%) had severe sepsis associated with bacteremia; 27 (20%) had MRSA bacteremia; 105 (82%) received effective therapy in the first 48 hours; and 102 (80%) survived hospitalization. Of the 125 patients, 92 (74%) (3 patients were lost to follow-up) survived for 30 days after the onset of bacteremia.
Table 1 shows the characteristics of 128 patients with S aureus stratified by oxacillin sodium resistance. Patients with MRSA bacteremia were significantly older; more likely to have a history of MRSA colonization or infection (RR, 4.1), to have been hospitalized in the last year (RR, 2.4), to have been admitted from another facility (RR, 2.3), and to have a nosocomial infection (RR, 2.4); and less likely to use injection drugs (RR, 0.3) compared with those patients with methicillin-sensitive S aureus (MSSA) bacteremia. No difference was detected in patients with severe sepsis or in patients with MRSA bacteremia who survived compared with patients with MSSA bacteremia. However, patients with MRSA bacteremia were significantly less likely to receive effective therapy during the first 48 hours of infection compared with patients with MSSA bacteremia (P<.01).
When the cohort population was stratified by nosocomial or community-acquired bacteremia, the results changed significantly (Table 2). Among patients with nosocomial bacteremia (n=74), a history of MRSA infection or colonization, admission from another facility, and rapidly or ultimately fatal underlying disease were significant risk factors for MRSA bacteremia. Further analysis by logistic regression revealed that a history of MRSA infection or colonization was the only independent risk factor (data not shown). Among patients with community-acquired bacteremia (n=54), these factors were not significant predictors of MRSA bacteremia.
Patients who died within 30 days of bacteremia were significantly older and more likely to be white; to have an underlying illness with a poor prognosis; to have a nosocomial infection; and to have severe sepsis associated with the bacteremia compared with those patients who survived (Table 3). The RR of death in patients given ineffective empiric therapy was 0.82 (95% confidence interval [CI], 0.36-1.88). The RR was similar when adjusted for severe sepsis (RR, 0.66; 95% CI, 0.37-1.77), age older than 65 years (RR, 0.65; 95% CI, 0.28-1.47), race (RR, 0.76; 95% CI, 0.34-1.76), nosocomial infection (RR, 0.74; 95% CI, 0.43-1.56), and poor prognosis (RR, 0.84; 95% CI, 0.37-1.88), indicating that this lack of effect was not confounded by these variables. In addition, there was no evidence of interaction with these variables. In the subset of patients with MRSA bacteremia (n=38), the RR of death in patients given ineffective empiric therapy was 0.58 (95% CI, 0.22-1.46).
Given the emergence of vancomycin resistance in gram-positive organisms,1 including the recent reports5,6 of vancomycin resistance in S aureus, the use of vancomycin clearly needs to be optimized. Vancomycin often is used empirically, and physicians often are reluctant to change antibiotics once a patient's symptoms have responded to a specific antibiotic.7 Ideally, the empiric use of vancomycin should be guided by the risk of MRSA infection in an individual patient, and thus, information on predictors of MRSA bacteremia in the local population should be incorporated into policies for the restriction of vancomycin use. Prior MRSA colonization was the strongest risk factor for MRSA bacteremias in this study of patients with S aureus bacteremia, although the RR was only significant in patients with nosocomial infection. This significance may be explained, in part, by the smaller sample size of community-acquired pneumonia. Among patients with nosocomial infections, hospitalization during the past year and admission from another facility also were risk factors for MRSA bacteremia; however, they were not independent risk factors when adjusted for a history of MRSA colonization.
A history of MRSA colonization or infection has not been studied previously as a predictor or risk factor for MRSA bacteremia. Mylotte and colleagues8 reviewed all clinically significant S aureus bacteremias from another Veterans Affairs facility from 1977 to 1985 and found that all MRSA bacteremias were nosocomial. Romero-Vivas and colleagues9 reviewed patients with nosocomial S aureus bacteremias and found that patients with MRSA bacteremia were significantly older, had significantly longer hospitalization, were administered more antibiotics, and underwent more surgical procedures than patients with MSSA bacteremia.
The MRSA bacteremias were less common in patients with community-acquired bacteremias. Patients who use injection drugs and are diagnosed as having community-acquired S aureus bacteremia at admission to the hospital were actually at a decreased risk of MRSA bacteremia. However, this finding may be geographically distinct. Another article10 reported injection drug use as a risk factor for MRSA bacteremia. Although a history of MRSA colonization was not a significant risk factor for patients with community-acquired MRSA bacteremia, these patients were more likely to have a history of MRSA colonization or infection compared with patients with community-acquired MSSA bacteremia.
This study did not find a significant difference in patient survival in those patients with S aureus bacteremia who did not receive any effective therapy for the first 48 hours compared with patients with S aureus bacteremia who received effective therapy during that time. This association was consistent when adjusted for nosocomial infection or severe sepsis and also remained when only patients with MRSA bacteremia were examined. In a minority of patients with S aureus bacteremia, the infection progresses to sepsis, and thus, initially untreated infections might not necessarily lead to a poorer outcome. The lack of effect also could be due to the type of supportive care received by patients. However, given the low mortality and small sample size, the CIs are relatively wide and the RR of death due to ineffective therapy could be as high as 1.46 in patients with MRSA bacteremia.
Given the emergence of vancomycin resistance in gram-positive organisms, vancomycin use needs to be carefully monitored, and thus, policies for the empiric use of vancomycin need to be developed. The optimal policy would identify patients at high risk for MRSA bacteremia and allow empiric therapy using vancomycin in this group. In the study population, which did not include patients with neutropenia, patients with a history of MRSA colonization were at highest risk of MRSA infection particularly when they had nosocomial infections. However, this finding also can be applied to patients with neutropenia. Vancomycin should be included in the initial antibiotic regimen of patients with febrile neutropenia and known MRSA colonization.11 Despite the lack of association between survival and effective empiric therapy, it seems prudent that a patient with a history of MRSA colonization should receive empiric therapy with vancomycin when an infection, which could be caused by S aureus, is suspected.
Accepted for publication August 3, 1999.
Corresponding author: Mary-Claire Roghmann, MD, Veterans Affairs Maryland Health Care System, 10 N Greene St, Room 5D137 (BT111), Baltimore, MD 21201 (e-mail: email@example.com).