Creation of a Combination Antibiogram for Pseudomonas aeruginosa in a Pediatric Intensive Care Unit

Abstract

OBJECTIVE

This study describes the creation of a combination antibiogram directed toward Pseudomonas aeruginosa to determine the most appropriate empiric antimicrobial regimen(s).

METHODS

P aeruginosa isolates were collected from all sites between January 2013 and December 2017 for patients admitted to the PICU. Patients with cystic fibrosis and isolates from the same site and susceptibility pattern obtained within 30 days were excluded. β-Lactam susceptibilities were determined and compared with the addition of an aminoglycoside or fluroquinolone and summarized in a combination antibiogram.

RESULTS

One hundred ninety-nine P aeruginosa isolates were included for analysis. The addition of a second agent to piperacillin-tazobactam was shown to have the most significant improvement among the β-lactams, with 70% susceptibility as monotherapy and increases to above 90% with the addition of an aminoglycoside or fluroquinolone. The addition of an aminoglycoside or fluroquinolone to cefepime and meropenem increased coverage to above 95%. The addition of a second agent was likely to increase susceptibility of a monotherapy backbone; however, as the susceptibility of the first-line agent decreased, the susceptibility of the second agent needed to be higher to achieve a 95% coverage threshold.

CONCLUSIONS

Our results support use of a second agent to significantly improve the likelihood of appropriate empiric coverage of P aeruginosa. Use of a combination antibiogram may be more beneficial than a simple antibiogram for units with increasing resistance rates, or for coverage of specific resistant organisms.

Introduction

Rapid initiation of appropriate empiric antimicrobial coverage is necessary to decrease the risk of morbidity and mortality within pediatric sepsis, septic shock, community-acquired pneumonia, and ventilator-associated pneumonia.^1–3 Infections with Pseudomonas aeruginosa have high overall mortality rates, which are worsened when patients receive initial antimicrobial therapy for which the Pseudomonal isolate is not susceptible.⁴ Local antibiograms have been recommended to ensure that empiric antimicrobial selections are based on susceptibility patterns within a distinct patient population.⁵ P aeruginosa has been shown to have increased resistance to antimicrobials and recommendations related to combination antibiotic therapy for P aeruginosa have been advocated, specifically in areas with monotherapy resistance rates ≥10% or in patients who have risk factors for resistant Gram-negative organisms.^6–9 This combination therapy typically includes an antipseudomonal β-lactam and a fluoroquinolone or aminoglycoside with activity against P aeruginosa.¹⁰

Standard local antibiograms are limited in their abilities to determine the interaction between susceptibilities for different antibiotics and overall susceptibility of these combinations. At many hospitals, the resistance rate of P aeruginosa to specific antibiotics can exceed 10%.¹¹ Studies suggest clinicians may be willing to accept the possibility of incorrect choices in up to 5% of the time for high-risk situations.^12,13 These high antibiotic resistance rates could lead to a greater chance of inappropriate antibiotic selection when standard antibiograms are used and cause clinicians caution when choosing a combination therapy. Combination antibiograms have been used and proposed in adult populations; however, this has not been detailed within pediatric populations.^10,14–16 The primary objective of this study was to describe the creation of a combination antibiogram directed toward P aeruginosa to determine the most appropriate empiric antimicrobial regimen for P aeruginosa. Secondary objectives were to determine the interaction between resistance patterns where monotherapy susceptibility rates may be known and simulate scenarios in which combination therapy may improve empiric antibiotic selection.

Materials and Methods

This was a retrospective cohort study that was approved by the Institutional Review Board at Nationwide Children's Hospital. We included patients admitted to the PICU from January 1, 2013, to December 31, 2017, who had a positive culture for P aeruginosa from any site obtained during PICU admission. We excluded cultures from patients with cystic fibrosis because these are often obtained for surveillance and frequently represent colonization. We also excluded cultures obtained within 30 days from the same site of infection and had the same susceptibility pattern.

P aeruginosa identification and susceptibilities were generated by automated VITEK 2-based Gram-negative cards (bioMerieux, Inc, Durham, NC). Interpretation of susceptibilities for piperacillin-tazobactam, cefepime, ceftazidime, meropenem, ciprofloxacin, gentamicin, tobramycin, and amikacin followed the guidelines of the Clinical and Laboratory Standards Institute's M100 series.⁹ P aeruginosa isolates with intermediate or resistant susceptibility to piperacillin-tazobactam, cefepime, ceftazidime, and meropenem were analyzed to determine if the addition of a second agent (ciprofloxacin, gentamicin, tobramycin, or amikacin) would improve the coverage of the isolate. Pearson χ² testing was used to determine if the addition of a second agent provided a statistically significant improvement in coverage using an α level of 0.05.

To estimate potential combination susceptibility rates where only monotherapy susceptibility rates were known, we used Monte Carlo simulation and standard probability theory formulas. Theoretical values for monotherapies were generated (100,000 replicates each for first-line fl and second-line [SL] mono-therapy) assuming FL monotherapy susceptibility rates ranging from 40% to 99% and SL monotherapy rates ranging from maximum (50, FL%–10%) up to 99%. Minimum values for the SL susceptibility rate were variable to ensure that when the SL therapy susceptibility rates were randomly generated, their rate would be not much lower than that of the FL therapy. We also simulated intersect values (FL ∩ SL, susceptibility rates based on using either FL or SL as monotherapy), constrained by the Bonferroni Inequality. Combination therapy probabilities were then calculated from the simulated data, using the standard probability formulas to solve for the probability of the union of 2 quantities (details in supplement). To summarize the results, FL and SL monotherapy susceptibilities were binned into 10% increments, and average susceptibility rates for each combination of FL and SL monotherapy strata were estimated, with 95% credible intervals.

Results

A total of 250 P aeruginosa isolates were included during the study period. After removing isolates from the same site within 30 days and isolates from cystic fibrosis patients, a total of 199 isolates were included in the final analysis. Culture sites included 150 (75.4%) cultures from the respiratory tract, 11 (5.5%) cultures from blood, 7 (3.5%) cultures from the urine, and 31 (15.6%) cultures from other sources.

Monotherapy results detailed susceptibility rate of 89%, 87%, 86%, and 79% for cefepime, ceftazidime, meropenem, and piperacillin-tazobactam, respectively (Table 1). The monotherapy susceptibilities for the combination agents were 97%, 89%, 89%, and 94% for amikacin, ciprofloxacin, gentamicin, and tobramycin (Table 1). There was an improvement in susceptibilities after 2013 for all 4 antipseudomonal β-lactams and after this improvement susceptibilities were maintained at higher percentages from 2014 to 2017 (Figure).

Table 1.

Susceptibility of Monotherapy and Combination Therapy for Pseudomonas aeruginosa Isolates (N = 199)

Antibiotics	% Susceptible
	Monotherapy	+ Amikacin*	+ Ciprofloxacin*	+ Gentamicin*	+ Tobramycin*
Cefepime	89	98	96	95	97
Ceftazidime	87	99	97	96	97
Meropenem	86	97	94	95	97
Piperacillin-tazobactam	79	98	94	94	97
Amikacin	97	NA	NA	NA	NA
Ciprofloxacin	89	NA	NA	NA	NA
Gentamicin	89	NA	NA	NA	NA
Tobramycin	94	NA	NA	NA	NA

NA: Not assessed

^* p < 0.05 for the different between monotherapy versus combination therapy.

Figure.

Monotherapy susceptibilities for antipseudomonal β-lactams over study period.^*

The results of the combination antibiogram creation are detailed in Table 1. In total, all combinations resulted in a significant increase in coverage when a second agent was added to the β-lactam backbone. The addition of amikacin or tobramycin to each of the antipseudomonal β-lactams resulted in a susceptibility of 95% or greater for each combination, with ceftazidime + amikacin being the combination with the highest overall susceptibility (99%). The addition of ciprofloxacin to meropenem or piperacillin-tazobactam still resulted in a significant increase in susceptibility when compared with single agent β-lactam but failed to reach 95% susceptibility threshold attaining a 94% susceptibility for each combination. The same is true for the combination of piperacillin-tazobactam and gentamicin.

Table 2 provides simulation results to estimate the combination therapy susceptibility rates when only monotherapy susceptibilities are known. We found that combination therapy susceptibility is relatively high (at least 76%), even when monotherapy susceptibilities are low (40%–50%) for each simulated antibiotic. Moreover, the relative improvement in susceptibility rates when adding a second antipseudomonal antibiotic is greater for lower monotherapy susceptibility rates. Adding a SL antibiotic agent with 90% to 100% susceptibility was able to improve antimicrobial susceptibility to >95% for each of the susceptibility categories of the β-lactam backbone.

Table 2.

Estimated Combination Therapy Susceptibility Rates Based on Monotherapy Susceptibility Rates (Percentage), With 95% Credible Intervals in parentheses ^*

	Second-Line Monotherapy Susceptibility, %
	50%–60%	60%–70%	70%–80%	80%–90%	90%–100%
First-line monotherapy susceptibility
40%–50%	76.75 (54.67–98.18)	82.77 (64.28–99.15)	87.36 (73.96–99.34)	92.55 (83.13–99.59)	97.29 (92.11–99.90)
50%–60%	78.55 (57.14–98.78)	82.26 (64.04–99.02)	87.41 (73.52–99.43)	92.60 (83.27–99.64)	97.29 (92.15–99.89)
60%–70%	82.17 (63.08–99.18)	83.49 (66.53–99.14)	87.48 (73.53–99.46)	92.49 (83.0–99.66)	97.24 (92.02–99.91)
70%–80%	NA	86.49 (72.97–99.30)	88.23 (75.75–99.46)	92.50 (83.09–99.56)	97.34 (92.07–99.90)
80%–90%	NA	NA	91.94 (82.54–99.62)	93.41 (85.24–99.67)	97.27 (92.05–99.91)
90%–100%	NA	NA	NA	96.79 (81.75–99.87)	98.16 (94.07–99.93)

NA: Not assessed

^* Example: If the first-line monotherapy susceptibility is known to be between 50% and 60% and the second-line monotherapy susceptibility is known to be between 70% and 80%, then the anticipated combination therapy susceptibility would be around 87.4%.

Discussion

The results of this study suggest that this tool could be used to improve empiric antimicrobial selections toward P aeruginosa in the PICU. Frequently, empiric antimicrobial therapy is based on the knowledge of single agent antibiograms and clinicians are unable to determine the overall impact of the addition of a second agent. As previously discussed, P aeruginosa can display high rates of resistance within the PICU, making the appropriate empiric antimicrobial therapy important for patient outcomes.^2,11 The use of a combination antibiogram has been described within adult intensive care units and large database studies; however, to the authors' awareness, this is the first description of a combination antibiogram within the pediatric population.^10,16

The 2020 Surviving Sepsis Guidelines for Septic Shock and Sepsis-Associated Organ Dysfunction in Children suggest considering the addition of a second agent at a threshold of 10% resistance for local antimicrobial resistance for patients with septic shock.⁵ Secondarily, the Infectious Diseases Society of America adult hospital-acquired and ventilator-associated pneumonia guidelines suggest a target threshold of ≥95% for empiric coverage when treating this patient population.^12,13,17 Within the adult population, the creation of a combination antibiogram assisted clinicians in selecting the most appropriate empiric antimicrobial regimen and found that the addition of a second agent increased the portion of P aeruginosa with adequate coverage.¹⁰ We advocate that a combination antibiogram should be considered to ensure the selection of the most appropriate antimicrobial regimen within the pediatric population, especially when resistance rates to β-lactams meet the above thresholds. Within our results, all our β-lactam backbone options met the resistance rate threshold for combination antibiotic therapy as described within the Surviving Sepsis guidelines, and all combinations improved coverage above 90%, with a majority meeting the recommended hospital-acquired and ventilator-associated pneumonia threshold.

In many institutions, empiric antimicrobial selection is based on antibiograms that only report susceptibilities of single agents, which can have limitations if attempting to use combination therapy to improve appropriateness of antibiotics. Secondarily, laboratories typically only share aggregate susceptibility data with individual units and combination antibiograms require the individual evaluation of cultures to determine susceptibility. This may be due to the overall limited number of cultures that are within the unit or limitations in sharing abilities of the institution's laboratory. The simulations within our study can be used to help institutions that report single agent susceptibility antibiograms determine which of their combinations will have the largest improvement in theoretical appropriateness for empiric antimicrobial therapy. Within our simulations, we saw that many combinations reached 90% susceptibility; however, as the β-lactam backbone susceptibility decreased, the SL monotherapy susceptibilities needed to be higher to reach this same threshold. Our study provides an opportunity for units to use their aggregate data to determine their empiric combination therapy to reach suggested thresholds. From our analysis, a second agent with at least 80% susceptibility should be added to the β-lactam backbone, to reach a target of 90%, and a second agent with at least 90% susceptibility should be added to reach a target of 95%. Publishing a yearly antibiogram for an institution or unit is common practice throughout hospitals, and these values can be used to establish a regimen that will meet targets proposed.^5,17

Amikacin and tobramycin provided the most substantial increases in coverage when added to an antipseudomonal β-lactam backbone within this patient population. This is likely due to the improved coverage of P aeruginosa that these agents provide in comparison with fluroquinolones and gentamicin.¹⁸ Fluroquinolone use within adult populations has been recommended against due to high rates of resistance and the lack of additional coverage when added to antipseudomonal β-lactam; however, we did not see this effect in our pediatric population.^19,20 This is likely due to the minimal usage of fluroquinolones within the pediatric population due to safety concerns.²¹ Secondarily, there may be overlapping-resistance between fluroquinolones and β-lactams due to the expression of efflux pumps for each of these antibiotic classes in P aeruginosa.²² Recent literature shows that patients admitted to the PICU may be at higher risk of acute kidney injury given certain concomitant antimicrobials (vancomycin plus piperacillin-tazobactam).²³ Clinicians must balance the mortality and morbidity benefits of adequately treating an infection with the consideration of medication toxicity. The substitution of a fluroquinolone for an aminoglyco-side may be a viable option in patients prescribed these concomitant antimicrobials given the maintenance high susceptibilities when combination therapy is needed to avoid excess exposure to nephrotoxic agents.

There are several limitations with our findings. First being its retrospective and single center design, which limits is generalization to other PICUs. However, this process can be translated to other units to determine their local susceptibilities, and our simulation data can be applied to other units in a more generalized approach. We saw a significant increase in our susceptibilities from 2013 to 2014, which was likely due to the opening of a new pulmonary intensive care unit and the transition of chronic tracheostomy and mechanically ventilated patients out of this data set. Second, this information is only related to the in vitro susceptibilities and is not related to clinical outcomes of patients, which can be impacted by multiple different factors.²⁴ We were unable to obtain levofloxacin susceptibilities for our study, which may be used more frequently in other institutions due to its favorable dosing regimens. Ciprofloxacin has been shown to have higher in vitro activity, but levofloxacin has comparatively higher concentrations with equivalent doses, making them comparative in most clinical circumstances.²⁵ Also, new β-lactam–β-lactamase inhibitors were not evaluated in this study; however, there is not currently enough data within the pediatric population, and these antibiotics are likely to be highly restrictive when available. Lastly, our β-lactam monotherapy regimens had relatively high susceptibility rates (79%–89%), and our combination therapy regimens had high susceptibility rates (89%–97%). Our specific combination antibiogram results may differ from units that have significantly lower susceptibilities for antimicrobials; however, the concepts explored in our simulation data would still apply to any unit so long as they were aware of their individualized susceptibility patterns.

Conclusion

The creation and use of a combination antibiogram directed at P aeruginosa can help PICUs determine optimal local empiric antimicrobial coverage, especially when empiric monotherapy may have susceptibility rates of ≤90%. If patient level data are unavailable for a unit, estimations can be made to determine the most appropriate combination therapy based on our simulations of the susceptibilities of individual antibiotics.

ABBREVIATIONS

FL

first line

PICU

pediatric intensive care unit

SL

second line