Multiple Antibiotic–Resistant Pseudomonas aeruginosa and Lung Function Decline in Patients with Cystic Fibrosis

Abstract

Background

The goal of this study was to determine the association of multiple antibiotic– resistant Pseudomonas aeruginosa (MARPA) acquisition with lung function decline in patients with cystic fibrosis (CF).

Methods

Using data from Epidemiologic Study of Cystic Fibrosis (ESCF), we identified patients with spirometry data and MARPA, defined as PA (1) resistant to gentamicin and either tobramycin or amikacin and (2) resistant to ≥1 antipseudomonal beta lactam. MARPA had to be detected in a respiratory culture after ≥ 2 years of PA-positive but MARPA-negative respiratory cultures. Multivariable piecewise linear regression was performed to model the annual rate of decline in forced expiratory volume in 1 second (FEV₁) % predicted 2 calendar years before and after the index year of MARPA detection, adjusting for patient characteristics and CF therapies.

Results

In total, 4,349 patients with chronic PA and adequate PFT data were identified; 1,111 subsequently developed MARPA, while 3,238 patients were PA positive but MARPA negative. Compared with patients who did not acquire MARPA, MARPA-positive patients had lower FEV₁ and received more oral (p<0.013) and inhaled p<0.001) antibiotic therapy. Mean FEV₁ decline did not change significantly after MARPA detection (−2.22 % predicted/year before detection and −2.43 after, p=0.45). There was no relationship between persistent infection or FEV₁ quartile and FEV₁ decline.

Conclusions

Newly detected MARPA was not associated with a significant change in the rate of FEV₁ decline. These results suggest that MARPA is more likely to be a marker of more severe disease and more intensive therapy, and less likely to be contributing independently to more rapid lung function decline.

Introduction

Chronic bacterial endobronchial infection is a fundamental feature of cystic fibrosis (CF) lung disease [1]. Antibiotics are cornerstones of CF therapy, but their use is associated with the detection of resistant organisms. Although patients with CF can be infected with a variety of organisms, the most common organism, especially in older patients, is Pseudomonas aeruginosa (PA) [2]. The prevalence of multiple antibiotic–resistant PA (MARPA) has been shown to have increased 40% from 1998-2002 [2]. The relationship between MARPA infection and lung function in patients with CF is unclear. Block, et al identified risk factors for pulmonary exacerbations in CF patients infected with multi-resistant bacteria, but they did not compare this group with patients who did not harbor MARPA [3]. Davies, et al studied the prevalence and persistence of MARPA infection in the pediatric CF population, but they did not analyze outcomes associated with MARPA infection [4]. In one study, the presence of MARPA was associated with more rapid decline in lung function, but this was a single-center study of adult patients with CF performed at a site whose patient population also included many patients referred for lung transplantation [5]. These features limit the ability to generalize their results to the CF population at large. More importantly, the study was not designed to assess the impact of incident detection of MARPA on subsequent lung function. Because of the importance of PA infection in CF and concerns about antibiotic resistance, determining whether incident MARPA is associated with lung function decline would be of value.

It has been demonstrated that the decline in lung function seen in CF patients is associated with pulmonary exacerbations [6, 7]. Exacerbations are frequently treated with intravenous antibiotics directed at the organisms detected by respiratory culture, most commonly PA [1]. The increasing emergence of MARPA raises the concern that antibiotic treatment of infections may be less effective, leading to accelerated loss of lung function. On the other hand, there is increasing evidence of poor correlation between resistance patterns and response to antibiotic therapy for pulmonary exacerbation [8-10]. This would imply that incident MARPA would not be a risk factor for accelerated lung function decline. To gain further insight into this issue, we used data from the Epidemiologic Study of Cystic Fibrosis (ESCF), a large observational study of patients with CF in North America [11], to assess the effect of incident detection of MARPA on lung function decline.

Methods

Study Design

We performed an analysis of data from the ESCF, a longitudinal cohort study. Data collected from 1994 to 2005 were used for this analysis. Details of the design and implementation of ESCF have previously been described [11]. Data were reported from every patient encounter and included pulmonary function tests (PFTs), anthropometric measures, clinical findings, respiratory tract cultures, other medical conditions (e.g., asthma), and use of therapies. Informed consent was obtained according to decisions by a local or central human subjects review board (Copernicus Group tracking number OVA1-03-008).

Cohort Definition

To be included in the analysis, patients had to be at least 6 years old and have at least one respiratory tract culture per year for a consecutive 5-year period. A consistent definition was used from 1994 to 2005, in which study sites were instructed to indicate MARPA if PA was found to meet both of the following criteria: (1) resistant to gentamicin and either tobramycin or amikacin and (2) resistant to one or more of the following antipseudomonal beta lactam antibiotics: ticarcillin, pipercillin, ceftazidime, aztreonam, and imipenem. Quinolones were not considered. For those laboratories where gentamicin was not tested, gentamicin resistance was assumed. Participants who had MARPA detected at any time during the observation period (1996–2003) and who had cultures negative for MARPA in the prior 2 years were considered incident MARPA cases (MARPA+). The date that MARPA was first reported was considered the date point for the subsequent analysis of pulmonary function decline. Each patient had to have a minimum of 3 PFTs spanning a minimum time of 6 months during the 2 years preceding and 2 years following the index year, resulting in a total period of 5 years (Figure 1). To account for short-term changes in pulmonary function that could result if treatment were immediately changed following a first-time MARPA-positive culture, we defined an exclusion period composed of the 90 days before and after the index date from which PFT results were excluded from the primary analysis. PFTs obtained during this exclusion period did not count towards the minimum of 3 required to be in the analysis.

Fig. 1.

Culture and PFT criteria for inclusion into the analysis cohort.

To allow comparisons between patients who acquired MARPA and those who did not, we also analyzed a comparison group of MARPA-negative patients composed of patients ≥6 years old who had PA but did not have any MARPA-positive cultures during the a 5-year period that satisfied the PFT criteria described above. The index date for these patients was defined as the date of the culture closest to July 1 of the index year. For MARPA-negative patients who were eligible for multiple index years, the most central index year was chosen if applicable; if 2 years were equally central, one was chosen at random as the index year.

Statistical Analysis Plan

The primary outcome measure for the analysis was the forced expiratory volume in 1 second (FEV₁). Values for FEV₁ % predicted were calculated using reference equations from Wang et al for males through age 17 years and for females through age 15 years and from Hankinson et al for patients over these ages [12, 13]. Spirometry data were matched with the data from the closest encounter within 14 days in order to obtain encounter-level covariates. If no encounters were present within the 14-day window, data from the most recent encounter preceding the PFT were used to obtain the covariates.

A previous ESCF analysis showed that sites whose median FEV₁ was in the highest quartile tended to use more antibiotics, especially intravenous antibiotics, compared with sites in the lowest quartile [14]. This same analysis also showed that these upper-quartile sites also had more MARPA. Since increased antibiotic use was associated with an increased likelihood of antibiotic resistance in this previous ESCF analysis and in other studies [15, 16], we stratified sites based on median FEV₁ % predicted at each site using the same approach as previously described [14] and studied the impact of MARPA detection on FEV₁ decline by site quartile.

Standard descriptive statistics were used to characterize the MARPA-positive and negative patients, with t tests and chi-square tests used to compare the two groups during the time periods pre- and post-index culture and during the exclusion period. Repeated-measures multivariable piecewise linear regression models were generated to examine the rate of decline in FEV₁ % predicted before and after the exclusion period. Patient-level covariates included in the models were gender and age at index culture, while encounter-level covariates included in the models were gender, age at index culture date, therapies (oral and inhaled antibiotics, intravenous antibiotic–treated exacerbations, oral and inhaled bronchodilators, oral and inhaled corticosteroids, mast cell stabilizers, oral, enteral, and parenteral supplements, airway clearance techniques, pancreatic enzymes, insulin, oxygen, dornase alfa, diagnosis of allergic bronchopulmonary aspergillosis, elevated liver function tests, and presence of other bacteria in most recent respiratory tract culture (PA, Staphylococcus aureus [SA], Haemophilus, Stenotrophomonas maltophilia [SM], Burkholderia cepacia, other gram-negative organisms, Aspergillus, and Candida). All variables were assessed as time-varying covariates, except gender and age at index culture, which were time independent. A two-sided p value <0.05 was considered statistically significant.

Results

Of the 32,585 patients enrolled in ESCF from 1994–2005, 13,958 patients (comprising 51,374 patient-years) had at least one respiratory-tract culture per year in each of the 5 calendar years associated with any of the index years (1996–2003). Of these patients, 4,349 met the remainder of the culture and PFT criteria outlined in the Methods; 1,111 patients had new MARPA detected (MARPA+), and 3,238 remained MARPA negative (Figure 2). The annual incidence of MARPA conversion in patients who were PA positive ranged from 6.2% to 9.2% (mean = 7.5%), and it did not change significantly over time (p = 0.72). Table 1 shows the clinical features of MARPA-positive patients compared with MARPA-negative patients at the index date. The mean age in both groups was similar at about 19 years, but there were significantly more young patients (age 6–11 years) in the MARPA-negative group (p <0.001, data not shown). MARPA-positive patients had mean FEV₁ values that were 10% predicted points significantly lower lung function than the MARPA-negative patients (p <0.001), a 15% relative reduction. They were also more likely to be treated with antibiotics by any route or receive other CF therapies, such as dornase alfa.

Fig. 2.

Derivation of the MARPA-positive and MARPA-negative cohorts.

Table 1.

Clinical features of the MARPA-positive and MARPA-negative cohorts. Unless otherwise noted, values in parentheses represent %. Aspergillus status refers to a positive culture and not the diagnosis of allergic bronchopulmonary aspergillosis.

Feature	MARPA Positive	MARPA Negative	p
n	1111	3238
Age at index culture (SD), y	19.7 (8.54)	19.1 (9.03)	0.064
Gender
Male	551 (50.9)	1273 (49.3)	0.40
Female	532 (49.1)	1307 (50.7)
Mean FEV1 % predicted (SD)	60.5 (23.8)	70.8 (25.3)	<0.001
Mean Number of cultures/years (SD)	3.4 (1.6)	2.8 (1.3)	<0.001
Mean Body Mass Index Z Score (SD)	–0.632 (1.15)	–0.555 (1.44)	0.121
Other organisms
MRSA	55 (8.8)	68 (6.6)	0.10
S. maltophilia	60 (5.4)	71 (4.1)	0.093
B. cepacia	23 (2.1)	38 (2.2)	0.86
Aspergillus	144 (13.0)	181 (0.3)	0.031
Treatment
Intravenous antibiotic-treated exacerbations	505 (45.5)	592 (18.3)	<0.001
Oral antibiotics	155 (14.3)	293 (11.4)	0.013
Inhaled antibiotics	481 (44.4)	795 (30.8)	<0.001
Oral corticosteroids	222 (20.5)	297 (11.5)	<0.001
Inhaled corticosteroids	486 (44.9)	935 (36.2)	<0.001
Dornase alfa	844 (76.0)	2173 (67.1)	<0.001
Supplemental oxygen	133 (12.3%)	125 (4.8%)	<0.001
FEV1, forced expiratory volume in 1 second; MARPA, multiple antibiotic-resistant Pseudomonas aeruginosa, MRSA, methicillin-resistant Staphylococcus aureus; SD, standard deviation.

The results of our primary outcome measure analysis of FEV₁ decline are shown in Figure 3. FEV₁ decline in MARPA-positive patients was −2.22 % predicted/year before the index event of MARPA detection and −2.43 % predicted/year after detection, a difference in slope that was not significant (p = 0.45). To control for the possibility that new MARPA detection was associated with an acute change in FEV₁ that may have been associated with more aggressive therapy, we also analyzed a model that allowed for discontinuities around the index event. We found no difference in our results using a discontinuous model. MARPA detection tended to be a transient event for some patients. In 36.3% of patients, subsequent cultures over the next 2 years were all negative for MARPA. To account for the possibility that transient MARPA patients were included in our analysis of persistent MARPA infection, we analyzed FEV1 decline in transient MARPA and persistent MARPA patients (Table 2). We found no significant difference in FEV₁ decline between these two groups.

Fig. 3.

Relationship between incident MARPA detection and FEV₁ decline. The mean rates of FEV₁ decline were estimated for the 2 years prior to and after the index event of MARPA detection in the MARPA+ group. FEV₁ represents the mean value up to and including the index event. Error bars represent ± 1 SD.

Table 2.

FEV1 decline in MARPA negative, transient MARPA, and persistent MARPA patients.

Patient Group	FEV1 slope before exclusion period (% predicted/year)	FEV1 slope after exclusion period (% predicted/year)	P
MARPA negative	−1.77	.2.45	<0.001
Transient MARPA	−2.12	−2.41	0.349
Persistent MARPA	−2.64	−2.62	0.977

Use of all CF therapies increased similarly over time for both groups, but the use of enteral supplements and oxygen increased even more in the MARPA-positive group compared with the MARPA-negative group (Table 3). Enteral supplement use increased from 17% to 25% in MARPA-positive patients compared with 9% to 15% in the MARPA-negative group (p = 0.017). The percentage of patients who ever received supplemental oxygen therapy increased from 17% to 29% in the MARPA-positive group vs 8% to 16% in the MARPA-negative group (p = 0.002)

Table 3.

Use of other therapies before and after the index date

Variable	Group	Before Exclusion Period, Mean (SE)	After Exclusion Period, Mean (SE)	p*	p*
Oral antibiotics (percent of encounters)	MARPA+ MARPA-	5.87% (0.4%) 7.37% (0.33%)	18.43% (0.97%) 18.18% (0.57%)	<0.001 <0.001	0.14
Inhaled antibiotics (percent of encounters)	MARPA+ MARPA-	22.46% (0.76%) 18.80% (0.45%)	39.54% (1.05%) 36.75% (0.65%)	<0.001 <0.001	0.44
Oral corticosteroids (percent of encounters)	MARPA+ MARPA-	13.86% (0.82%) 10.49% (0.46%)	15.89% (0.92%) 11.98% (0.50%)	0.016 0.002	0.58
Enteral supplements (percent of encounters)	MARPA+ MARPA-	12.19% (1.01%) 7.41% (0.51%)	18.69% (1.18%) 11.85% (0.64%)	<0.001 <0.001	0.051
Insulin/oral hypoglycemic (percent of encounters)	MARPA+ MARPA-	0.12% (0.05%) 0.19% (0.04%)	4.08% (0.53%) 3.51% (0.33%)	<0.001 <0.001	0.28
Dornase alfa (percent of encounters)	MARPA+ MARPA-	75.70% (1.18%) 65.62% (0.83%)	80.75% (1.06%) 74.65% (0.75%)	<0.001 <0.001	0.002
Oxygen (percent of patients ever)	MARPA+ MARPA-	17.37% (1.14%) 8.09% (0.48%)	29.43% (1.37%) 15.66% (0.64%)	<0.001 <0.001	0.002

SE, standard error.

^*Compares before to after within MARPA group.

^*Compares difference between MARPA groups.

The assessment of MARPA by FEV₁ quartile described in the Methods revealed that the prevalence of MARPA was similar between the upper and lower quartile sites (25.8% vs 25.7%, p = 0.96), with no relationship between FEV₁ quartile and FEV₁ decline before and after MARPA detection, regardless of age and baseline disease stage.

Discussion

The development of antibiotic-resistant bacteria is of concern to CF clinicians and researchers [17]. In this analysis of a large observational study of patients with CF, new MARPA detection was not associated with a change in FEV₁ decline. Analyses based on persistence of infection and site FEV₁ quartile also showed a lack of association between MARPA and FEV₁ decline.

It is difficult to compare the results of our study with those of other studies of MARPA, because of the varying definition of multiple antibiotic resistance. The US CF Foundation guidelines define MARPA as resistance to all agents tested in two or more of the following antibiotic classes: beta-lactams, aminoglycosides, and quinolones [17]. Our definition required aminoglycoside and beta-lactam resistance, but did not require resistance to all agents tested. Not all authors have applied the US CFF definition, and even the CF Foundation (CFF) Patient Registry only began collecting data using this definition in 2004. Similar to our findings with the ESCF definition of MARPA, the prevalence of CFF-defined MARPA has not changed from 2004 to the present [18]. In their analysis using the CFF Registry, Merlo et al used a very narrow definition of MARPA (resistance to at least two of the following antibiotics: tobramycin, ciprofloxacin, or meropenem) and found that the prevalence of MARPA rose steadily from 1998–2002 [2]. However, the Emerson et al. analysis of the CFF Registry compared a historical cohort of patients with CF in 1998 with a contemporary cohort in 2008 and found no change in MARPA prevalence after adjusting for age and baseline FEV₁ using the CFF definition [15]. Consistent with other studies of antibiotic resistance in CF [4, 16, 19-22], we found that antibiotic use was highly associated with the likelihood of acquiring MARPA.

Similar to the pattern observed for SM and methicillin-resistant SA [19, 20, 22], we found that MARPA infection was frequently a transient phenomenon, and only a small minority of patients remained consistently MARPA positive after the initial detection. Possible explanations for this finding include reversion of antibiotic sensitivity after selective pressure of antibiotic therapy is gone, acquisition of a new PA strain, or simple failure to recover MARPA consistently from respiratory cultures. We have no information on the sensitivity and reproducibility of the MARPA assays used at the various sites participating in ESCF. Regardless of the reasons for transient MARPA detection, we observed no effect of new MARPA detection on FEV₁ decline, even in chronically infected patients.

Only one other study has analyzed the impact of MARPA on clinical outcomes, and it was performed at a single adult CF center that also served as a referral site for lung transplantation. Letchzin et al reported that the presence of MARPA was associated with a higher risk of death, a greater likelihood of need for lung transplantation, and more rapid rate of FEV₁ decline compared with MARPA-negative adult CF patients, but their study was not designed to assess the impact of new MARPA detections on subsequent FEV₁ decline [5]. In our study, which included data from >200 pediatric and adult sites across North America, we also found that MARPA was associated with more rapid FEV₁ decline. We also have made the novel and important observation that MARPA acquisition does not result in a significant change in FEV₁ decline, suggesting that it is a marker for more advanced lung disease with PA present, rather than an independent contributor to worsening disease.

Our results are consistent with other data regarding CF microbiology and care. Although the acquisition of some antibiotic-resistant bacteria is associated with worsening FEV₁ decline [19, 23], this does not happen in all cases [20, 24]. The lack of association between in vitro antibiotic susceptibility and clinical response has been previously described [8-10]. Virulence of organisms may be determined by factors other than antibiotic resistance patterns. It may also be that the comprehensive treatment for CF exacerbation results in clinical improvement independent of the bactericidal impact of antibiotics used [8].

As with any study utilizing observational data, there are limitations to our analysis. There are no specific microbiology protocols in ESCF, so specimen processing may not be consistent across all sites [25]. However, the majority of patients are followed at CFF-accredited care centers, which are required to adhere to CFF guidelines for handling CF microbiology specimens and undergo regular site visits to ensure compliance with this practice. Because ESCF was initiated shortly before the wide availability of oral quinolones, the definition of MARPA was defined as resistance to both aminoglycosides and beta-lactams [11]. This makes comparison of our analysis to previous studies difficult, but other studies have also used definitions that differ from the current CFF guideline definition of MARPA [2]. The prevalence of ciprofloxacin resistance has not changed significantly in 2008 compared with 1995 [15]. This observation suggests that most MARPA has arisen because of development of aminoglycoside or beta-lactam resistance. We adjusted for a large number of potential confounders including all the available confounding therapies, co-morbid conditions, and other bacterial infections However, as with any observational study, there could be unmeasured confounders limiting the sensitivity of our analysis.

In summary, our analysis of ESCF data does not demonstrate an impact of MARPA acquisition on FEV₁ decline. Rather, MARPA appears to be a marker of an advanced stage of lung disease and the need for antibiotic therapy in patients with CF. Future studies should focus on molecular and genetic factors that may contribute to increased virulence of PA regardless of antibiotic resistance patterns, and on determining rational and clinically relevant treatment regimens for CF patients infected with MARPA, such as nutritional support, oxygen supplementation, and mucolytics.

Abbreviation List

CF

cystic fibrosis

CFF

Cystic Fibrosis Foundation

ESCF

Epidemiologic Study of Cystic Fibrosis

FEV₁

forced expiratory volume in 1 second

MARPA

multiple antibiotic–resistant Pseudomonas aeruginosa

PA

Pseudomonas aeruginosa

PFT

pulmonary function test

SM

Stenotrophomonas maltophilia