Optimizing Empiric Antibiotic Therapy in Patients with Severe β-Lactam Allergy

Abstract

Antibiotic selection is challenging in patients with severe β-lactam allergy due to declining reliability of alternate antibiotics. Organisms isolated from these patients may exhibit unique resistance phenotypes. The objective of this study was to determine which alternate antibiotics or combinations provide adequate empirical therapy for patients with β-lactam allergy who develop Gram-negative infections at our institution. We further sought to determine the effects of risk factors for drug resistance on empirical adequacy. A retrospective analysis was conducted for adult patients hospitalized from September 2009 to May 2010 who had a severe β-lactam allergy and a urine, blood, or respiratory culture positive for a Gram-negative organism and who met predefined criteria for infection. Patient characteristics, culture and susceptibility data, and predefined risk factors for antibiotic resistance were collected. Adequacies of β-lactam and alternate antibiotics were compared for all infections and selected subsets. The primary outcome was adequacy of each alternate antibiotic or combination for all infections. One hundred sixteen infections (40 pneumonias, 67 urinary tract infections, and 9 bacteremias) were identified. Single alternate agents were adequate less frequently than β-lactams and combination regimens. Only in cases without risk factors for resistance did single-agent regimens demonstrate acceptable adequacy rates; each factor conferred a doubling of risk for resistance. Resistance risk factors should be considered in selecting empirical antibiotics for Gram-negative pathogens in patients unable to take β-lactams due to severe allergy.

INTRODUCTION

Empirical antibiotic selection is important for optimizing therapy for bacterial infections. Morbidity and mortality increase for certain infections when empirical therapy is inadequate for the causative pathogens (1). Guidelines recommend consideration of local microbiologic data and patient risk factors for drug resistance when choosing empirical therapy (2). β-Lactam antimicrobials play a prominent role in the therapy of many community- and health care-associated infections (3, 4). Clinicians commonly encounter allergic intolerance to β-lactam antimicrobials, and this may limit therapeutic options, especially when the history includes a report of a severe reaction. Rates of resistance among Gram-negative pathogens to non-β-lactam antimicrobials are increasing, rendering these alternate agents less useful (5, 6). Studies to aid clinicians in selecting empirical therapy in this population are lacking.

An institutional antibiogram that tabulates laboratory susceptibilities of common pathogens to antibacterial drugs can be a useful tool in selecting empirical therapy (3, 7, 8). The specificity of antibiogram results for a given patient case can be increased by reporting on subsets of organisms from various body sites or from patients in various care locations. Most antibiogram tables report only single drug-organism susceptibility pairs, without assessing the adequacy of combinations of antibiotics (7). Patients with β-lactam allergy may have been exposed to other classes of antimicrobials more frequently than those without allergy and may be infected with pathogens exhibiting altered patterns of resistance.

Most β-lactam antibiotic agents are safe, but allergic reactions are not uncommon. In rare instances, these reactions may be severe (9, 10). Cross-allergenicity between β-lactam agents and other classes of antibiotics has been characterized (11). Antimicrobials unlikely to cross-react with β-lactams include fluoroquinolones, aminoglycosides, trimethoprim-sulfamethoxazole (11), tetracyclines, and the monobactam aztreonam (9, 10). Where severe allergy to one β-lactam agent is reported (e.g., anaphylaxis in response to penicillin), many clinicians avoid the use of related β-lactam agents, such as cephalosporins and carbapenems (9–11).

Recent reports have documented increasing rates of resistance to several non-β-lactam compounds among clinically encountered Gram-negative pathogens (5, 12–15). This may complicate empirical therapy for patients with serious bacterial infections who are unable to receive β-lactam drugs. Declining reliability of non-β-lactam antibiotic agents against Gram-negative pathogens in our health system over the past decade prompted an institution-specific study of this issue.

Several factors have been shown to be predictive of bacterial resistance in patients with serious infections. The American Thoracic Society and the Infectious Diseases Society of America mention several of these risks in their guideline documents on hospital-acquired pneumonia, health care-associated pneumonia, and ventilator-associated pneumonia (2). These include recent encounters in the health system, past antimicrobial therapy, and concurrent immunosuppression. Several published studies identified similar risk factors (6, 12–16).

Guidelines for the management of pharmacotherapy for patients with drug allergies have been published and outline several options for patients with a history of severe reactions to β-lactam drugs (17). These include (i) skin testing and then (ii) graded challenge or desensitization, or use of alternate (non-β-lactam) antibiotics. For treating serious infections in hospitalized patients, the time constraints of empirical therapy usually limit these options to the latter. The purpose of this study was to develop institution-specific recommendations for empirical therapy in patients with a severe β-lactam allergy who develop Gram-negative infections. The primary objective of this study was to determine which alternate antibiotics or antibiotic combinations provide the best empirical therapy for these infections. A secondary objective was to assess how selected risk factors for drug resistance influence the adequacy of antibiotics in this population.

(Part of this work was presented previously at the 50th Interscience Conference on Antimicrobial Agents and Chemotherapy, Boston, MA, 2010 [18].)

MATERIALS AND METHODS

We performed this retrospective analysis at Wake Forest Baptist Medical Center (WFBMC), an 885-bed academic hospital located in Winston-Salem, NC. Term definitions and the criteria for the diagnosis of infection are displayed in Tables 1 and 2, respectively. This study was approved by the WFBMC Institutional Review Board, and informed consent was waived.

Table 1.

Definitions of terms

Table 2.

Criteria for infection

Patient population.

Potential patients for inclusion were identified from a computer-generated report of blood, urine, and respiratory cultures. Adult patients hospitalized between 21 September 2009 and 21 March 2010 were included if they had a culture that grew a Gram-negative, aerobic organism; had susceptibility information available for that culture; had a history of severe β-lactam allergy; were ≥18 years old; and were admitted to a nonpediatric service. Patients were excluded if their cultures did not meet predefined criteria for infection (Table 2) or if the patient had previously been included during that hospital stay. Medical records were used to collect patient characteristics, including risk factors for Gram-negative antibacterial resistance.

Only the first culture of a patient's hospital stay was included, unless a second culture drawn within 24 h grew a different pathogen representing a separate infection by predefined criteria.

Microbiologic data.

Susceptibility data were recorded for the following antibiotics: aztreonam, ciprofloxacin, amikacin, gentamicin, tobramycin, trimethoprim-sulfamethoxazole, meropenem, piperacillin-tazobactam, and cefepime. Nitrofurantoin and tetracycline susceptibilities were recorded for urine isolates only. Susceptibilities were determined by an automated modified broth microdilution method (Microscan), using the Clinical and Laboratory Standards Institute susceptibility breakpoints (19). For the purposes of this study, susceptibility tests reported as “intermediate” were considered resistant.

Outcomes measured and statistical analysis.

The primary outcome was adequacy of each antibiotic or antibiotic combination for all infections. Adequacy was defined as in vitro activity against the Gram-negative isolate(s) from each culture (20, 21). Adequacy of combination regimens required at least one antibiotic in the regimen to exhibit in vitro activity. Secondary outcomes included adequacy based on type of infection and adequacy based on selected risk factors. Differing rates of adequacy between antibiotics and combinations were analyzed using the χ² or Fisher's exact test, as appropriate. A goal sample size of 120 patients was set to detect a difference of 15% between antibiotics or antibiotic combinations for the primary outcome in order to achieve 80% power and a two-sided alpha error of 0.05. To test the significance of the various risk factors for antimicrobial resistance (and thus lack of adequacy), a backward-stepping multivariate regression analysis was performed for three regimens (aztreonam, ciprofloxacin, and their combination). Included in this analysis were the risk factors listed in Table 1, as well as patient age, infection type, and the dichotomous variables of patient gender, intensive care unit (ICU) location, mechanical ventilation, presence of an indwelling bladder catheter, and presence of a central venous catheter. Factors were removed from the model if the P value was >0.10. We conducted a similar analysis examining the effect of the number of risk factors on the likelihood of resistance.

RESULTS

A total of 276 cultures were screened for inclusion, and 116 cultures met inclusion criteria. The major reasons for exclusion were failure to meet predefined criteria for infection and previous inclusion of the patient in the study. The 116 included cultures were from 109 patients; 5 patients were included twice due to second hospitalizations, and 2 had two cultures within 24 h that represented infections at differing sites (Fig. 1). Patient demographics are summarized in Table 3. Forty percent of infections were hospital acquired. One hundred thirty pathogens were isolated, and these included Escherichia coli (50 isolates [38.5%]), Pseudomonas species (24 isolates [18.5%]), Klebsiella species (20 isolates [15.4%]), Enterobacter species (12 isolates [9.2%]), and Proteus mirabilis (6 isolates [4.6%]). Other organisms comprised 13.8% of the total.

Fig 1.

Patients included and infection classification.

Table 3.

Patient characteristics and presence of risk factors (n = 109)

Characteristic	Valuea
Age (yr) (mean ± SD)	62 ± 18
Male	35
Hospital-acquired infection	40
ICU location at time of culture	36
At least 1 risk factor present	78
Presence of risk factor
Antibiotic use in prior 90 days	59
History of >2 days of hospitalization	38
Skilled nursing facility resident	18
Immunosuppressive therapy	13
Immunosuppressive disease	8
Recipient of home infusion therapy	7
End-stage renal disease	5

^aValues are percentages, unless indicated otherwise.

Primary outcomes.

Adequacies of antibiotics and antibiotic combinations for all infections are reported in Fig. 2 and 3. The adequacy provided by monotherapy with a traditional β-lactam exceeded the adequacies provided by all alternate agents except amikacin. Combining aztreonam and ciprofloxacin significantly improved the adequacy of coverage, to 85%, versus that of ciprofloxacin alone, while adding amikacin to ciprofloxacin or aztreonam significantly improved the adequacy of coverage, to 91% and 88%, respectively, compared to that of ciprofloxacin or aztreonam monotherapy. Compared to each other, these combinations did not differ significantly.

Fig 2.

Percent adequacies of antibiotics for all infections (n = 116). *, P < 0.05 versus aztreonam alone; †, P < 0.05 versus ciprofloxacin alone.

Fig 3.

Adequacies of antibiotic combinations for all infections (n = 116). Black bars represent adequacies achieved by the first antibiotics. White bars represent additional adequacies gained by adding the second agents. *, P < 0.05 versus aztreonam alone; †, P < 0.05 versus ciprofloxacin alone.

Secondary outcomes.

For the 40 pneumonias identified, aztreonam monotherapy was adequate in 55% of cases. Ciprofloxacin monotherapy demonstrated adequacy in 63% of cases. Adding amikacin to aztreonam and to ciprofloxacin resulted in improved adequacies compared to each monotherapy (75% and 86%, respectively). However, only the ciprofloxacin-amikacin regimen achieved a statistically significant improvement. The addition of ciprofloxacin to aztreonam resulted in 80% adequacy, which was also statistically significant compared to that of ciprofloxacin alone. Statistically significant differences were not noted between any two combination regimens (Fig. 4).

Fig 4.

Adequacies of antibiotics for pneumonia (n = 40). Black bars represent adequacies achieved by the first antibiotics. White bars represent additional adequacies gained by adding the second agents listed. *, P < 0.05 versus aztreonam alone; †, P < 0.05 versus ciprofloxacin alone.

For the 67 urinary tract infections identified, aztreonam resulted in the highest rate of adequacy (84%) among monotherapies. This was significantly higher than those of all other alternate monotherapies. The use of amikacin in combination significantly improved the adequacy of ciprofloxacin, to 94%. The addition of ciprofloxacin or amikacin to aztreonam improved adequacy to 85% or 94%, respectively, but these improvements were not statistically significant.

Adequacies of the combinations examined did not differ significantly from one another (Fig. 5).

Fig 5.

Adequacies of antibiotics for urinary tract infections (n = 67). Black bars represent adequacies achieved by the first antibiotics. White bars represent additional adequacies gained by adding the second agents listed. †, P < 0.05 versus ciprofloxacin alone.

With only nine primary bacteremias, this subgroup was not analyzed separately. We included these cases in the composite primary outcome and in the analysis of the influence of resistance risk factors.

Influence of risk factors.

When the effects of individual risk factors were analyzed using a backward-stepping logistic approach, several factors demonstrated significance, depending upon the antimicrobial regimen studied. These results are shown in Table 4 for three regimens: ciprofloxacin, aztreonam, and their combination. Note that for all three regimens, the odds ratio (OR) for hospital-acquired infection was significantly less than 1, suggesting that acquisition of infection in the hospital conferred less risk of resistance to these regimens than that with community-acquired infection.

Table 4.

Association between patient characteristics and antibiotic resistance

Parameter	OR (95% CI)a
	Ciprofloxacin resistance	Aztreonam resistance	Resistance to ciprofloxacin-aztreonam
Infection site
Blood (reference)	1.0	1.0	***
Respiratory	19.3 (1.6–229)†	57.0 (4.7–691)†
Urinary	27.2 (2.3–323)†	4.3 (0.4–44.0)
Hospital-acquired infection	0.19 (0.05–0.69)†	0.2 (0.05–0.85)†	0.19 (0.04–0.91)†
Presence of urinary catheter	2.7 (0.97–7.6)	***	***
Presence of central line	4.3 (1.1–16.0)†	3.3 (0.98–10.9)	4.6 (1.2–18.5)†
Receipt of mechanical ventilation	***	***	3.7 (0.97–13.8)
Receipt of immunosuppressive therapy	***	3.5 (0.86–14.7)	3.7 (0.85–16.2)
Receipt of antibiotics within prior 90 days	3.1 (1.2–8.1)†	***	***
Admission to hospital within last 90 days	***	5.1 (1.6–16.5)†	4.0 (1.2–13.7)†
Admission from skilled nursing facility or long-term care facility	5.3 (1.5–18.0)†	***	***

^a***, factor was dropped from model because the P value was >0.10; †, P < 0.05.

The relationship between the number of risk factors identified and adequacies of several regimens is shown graphically in Fig. 6. In patients lacking these risk factors (0 risks), the adequacy provided by aztreonam or ciprofloxacin was similar to that provided by their combination. Backward-stepping multivariate regression analysis described the cumulative relationship of the number of risk factors and decreasing adequacy. Each encountered risk factor conferred an approximately 2-fold risk of resistance (ciprofloxacin OR = 1.908 [95% confidence interval, 1.282 to 2.84] and P = 0.0015; aztreonam OR = 2.566 [1.597 to 4.123] and P < 0.0001; ciprofloxacin-aztreonam OR = 2.579 [1.560 to 4.262] and P = 0.0002).

Fig 6.

Adequacies of 4 regimens by number of risk factors for resistance (n = 116).

DISCUSSION

Patients with a history of serious β-lactam allergy represent a unique patient population. They are exposed to other classes of antibiotics more frequently than those without allergy, which could alter patterns of antimicrobial resistance. Their allergies also reduce the antibiotic options available for treatment. This study is the first to examine the issue of adequacy of empirical therapy within this unique group of patients. A goal of our study was to learn how to better care for infected patients with allergies to β-lactam drugs.

Most of the patients studied had a history of severe β-lactam allergy; however, patients were included who had a status of “reaction unknown.” This was done because of the “real-world,” clinical observation that when patients are unable to clarify their allergies, prescribers frequently avoid β-lactam antimicrobials altogether.

Several principles can be drawn from this investigation that could assist clinicians with choosing empirical therapy for patients at our institution. First, with the exception of amikacin, bacteriologic adequacy with non-β-lactam “alternate” antibacterials was less than that of traditional β-lactam drugs. This provides a strong motivation to ensure that a patient's allergy history is accurate and truly prevents the use of traditional β-lactam antibiotics.

Second, risk factors for resistance appeared to significantly influence the adequacy of various antibiotic regimens in this population. For each risk factor encountered, the risk of resistance (and thus for lack of adequacy) approximately doubled. We observed acceptable adequacies (above 90%) for single alternate antibiotic agents only in those patients in whom such risk factors were absent. This fact highlights the importance of assessing the patient's history when designing empirical therapy.

Third, combination therapies may be more toxic than monotherapies, especially if they include aminoglycosides. By clarifying the relative roles of the various antibiotics evaluated in this population, this investigation may better equip clinicians to weigh toxicity risks against the benefits of alternate antimicrobial regimens. Understanding situations in which these drugs are necessary and when they can be avoided can help with empirical decision making.

Lastly, this study demonstrates the importance of combination therapy within this population. With a few notable exceptions, single-agent, non-β-lactam empirical therapy in patients with severe β-lactam allergy proved inadequate more than 25% of the time. Combination regimens were found to have improved rates of adequacy compared to single agents. With the exception of aztreonam monotherapy for urinary tract infections, combination therapy is required to provide adequate empirical therapy if risk factors are present. We have incorporated these clinical principles into care algorithms and guidelines at our institution.

A somewhat unexpected finding in this study was the observation that infections acquired in the hospital were associated with less risk for resistance than those acquired elsewhere. Our study was not designed to identify reasons for this result; however, this may be related to the fact that use of ciprofloxacin and aztreonam is tightly controlled in our hospital and that inpatient exposure housewide to these antimicrobials has declined over the past 5 years. There are few restrictions on the usage of these agents outside our facility. Therefore, differences in rates of resistance could be related to differences in antibiotic exposure.

This investigation is not without limitations. Although the study was very close to achieving its target sample size, many differences between alternate therapies found to be nonsignificant may have been statistically significant if more patients were included. Due to the retrospective nature of the study, patient history data were limited to documentation in our electronic medical records. Data on some aspects of care, such as prior antibiotic use and previous colonization or infection with resistant organisms, may not have been available if not documented in the medical record. It should be noted that this was a study of the pathogens infecting patients admitted to our institution and that specific results may not be applicable to other settings.

In conclusion, this investigation demonstrates that compilation of patient and susceptibility data into a disease-specific combination antibiogram can be used to refine empirical antibiotic choices in a subpopulation of patients with severe β-lactam allergy. This approach demonstrated that, in our medical center, combination therapy improves empirical adequacy against infecting Gram-negative pathogens in this group of patients. The source of infection and risk factors for drug resistance should be considered when selecting empirical antibiotic therapy in this setting. This approach could be used to develop institution-specific treatment recommendations in other health care systems.