Pseudomonas aeruginosa antibiotic resistance in Australian cystic fibrosis centres
Abstract
Background and objective
In cystic fibrosis (CF), chronic Pseudomonas aeruginosa infection is associated with increased morbidity, antibiotic treatments and mortality. By linking Australian CF registry data with a national microbiological data set, we examined the association between where treatment was delivered, its intensity and P. aeruginosa antibiotic resistance.
Methods
Sputa were collected from paediatric and adult CF patients attending 18 Australian CF centres. P. aeruginosa antibiotic susceptibilities determined by local laboratories were correlated with clinical characteristics, treatment intensity and infection with strains commonly shared among Australian CF patients. Between-centre differences in treatment and antibiotic resistance were also compared.
Results
Large variations in antibiotic usage, maintenance treatment practices and multi-antibiotic resistant P. aeruginosa (MARPA) prevalence exist between Australian CF centres, although the overall proportions of MARPA isolates were similar in paediatric and adult centres (31% vs 35%, P = 0.29). Among paediatric centres, MARPA correlated with intravenous antibiotic usage and the Australian state where treatment was delivered, while azithromycin, reduced lung function and treating state predicted intravenous antibiotic usage. In adult centres, body mass index (BMI) and treating state were associated with MARPA, while intravenous antibiotic use was predicted by gender, BMI, dornase-alpha, azithromycin, lung function and treating state. In adults, P. aeruginosa strains AUST-01 and AUST-02 independently predicted intravenous antibiotic usage.
Conclusion
Increased treatment intensity in paediatric centres and the Australian state where treatment was received are both associated with greater risk of MARPA, but not worse clinical outcomes.
Abbreviations
-
- ACFDR
-
- Australian cystic Fibrosis Data Registry
-
- ACPinCF
-
- Australian Clonal Pseudomonas aeruginosa in cystic fibrosis
-
- BMI
-
- body mass index
-
- CDS
-
- Continuous Dichotomous Susceptibility
-
- CI
-
- confidence interval
-
- CF
-
- cystic fibrosis
-
- CLSI
-
- Clinical and Laboratory Standards Institute
-
- ERIC-PCR
-
- enterobacterial repetitive intergenic consensus polymerase chain reaction
-
- EUCAST
-
- European Committee on Antimicrobial Sensitivity Testing
-
- MARPA
-
- multi-antibiotic resistant Psedomonas aeruginosa
-
- FEV1
-
- forced expiratory volume in 1 s
-
- FVC
-
- forced vital capacity
-
- MIC
-
- minimum inhibitory concentration
Introduction
In cystic fibrosis (CF), chronic Pseudomonas aeruginosa infection is associated with accelerated pulmonary decline, reduced quality of life, increased treatment burden and poorer survival.1, 2 CF consensus statements therefore emphasize the importance of managing P. aeruginosa pulmonary infections.3, 4 However, commonly shared and multi-antibiotic resistant P. aeruginosa (MARPA) strains emerging in CF centres worldwide have complicated patient management and infection control policies.5
Utilizing information from national CF data registries allows between-centre comparisons and quality improvement programmes to benchmark against ‘high-performing’ centres.6-8 However, few studies have assessed P. aeruginosa treatment strategies using these national data sets. Here, data from the Australian Clonal Pseudomonas aeruginosa in CF (ACPinCF) study9 and the Australian CF Data Registry (ACFDR) were linked to examine differences in managing P. aeruginosa infection. Our aim was to examine the association between where treatment was delivered, its intensity and P. aeruginosa antibiotic resistance in the Australian CF population.
Methods
Patients, data and sample collection
The ACPinCF study involved 18 Australian CF centres and is described in detail elsewhere.9 Briefly, it examined the point prevalence, diversity and clinical impact of P. aeruginosa strains in Australian CF clinics, and included 2677 patients (1300 aged ≥18 years; centre size: 25–294), representing 91%, 89% and 90% of the paediatric, adult and total Australian CF population, respectively.10
Human ethics committees of all participating institutions approved the study. Participants provided written, informed consent, which for children included parents or legal guardians.
Each patient provided a single sputum specimen either at a routine clinic visit or during hospitalization between September 2007 and June 2010. Age, gender, inhaled maintenance therapies (tobramycin, colistin, dornase-alpha and hypertonic saline) and oral azithromycin in the previous 30 days; and number of courses and total days of intravenous antibiotics and outpatient clinic visits during the previous 12 months were recorded. The best forced expiratory volume in 1 s (FEV1) in the calendar year of sample collection (and forced vital capacity (FVC) and body mass index (BMI) from the same day as FEV1 measurement) reported to the ACFDR were also recorded. In children, age-adjusted pulmonary function prediction equations11, 12 and standard deviation z-scores for BMI were utilized.13
Laboratory testing
Sputum culture and antibiotic susceptibility testing were performed routinely by hospital laboratories. Six centres performed sensitivity testing on a ‘mixed’ colony sweep of culture plates, while in the remaining 12 centres individual ‘pure’ colonies were selected for testing.14 Disk diffusion sensitivity testing was applied by all centres, 12 used Clinical and Laboratory Standards Institute (CLSI) guidelines, five the Continuous Dichotomous Susceptibility (CDS) test and one applied European Committee on Antimicrobial Sensitivity Testing (EUCAST) methods and breakpoints15-17 (Supplementary Table S1). Three P. aeruginosa colonies representing different morphotypes from each specimen were selected locally and transported to the research laboratory to confirm their identity18 and for enterobacterial repetitive intergenic consensus polymerase chain reaction (ERIC-PCR) genotype testing.9
MARPA isolates were defined as being resistant to all antibiotics tested in two or more classes of anti-pseudomonal agents (aminoglycoside, β-lactam and fluoroquinolone antibiotics).19 Where anti-biograms for multiple isolates from the same sample existed (398/934 samples), the most resistant isolate was used. If an isolate's susceptibility to an antibiotic class was not tested, it was excluded from analysis of resistance to that antibiotic class, and if not already identified as a MARPA, from this category, too. Thus, 19 isolates were excluded from the MARPA analysis because of lacking data for aminoglycoside (n = 1), β-lactam (n = 2) and fluoroquinolone (n = 21) susceptibilities.
Shared P. aeruginosa strains
Shared P. aeruginosa strains had indistinguishable ERIC-PCR gel patterns from sputum isolates cultured from at least two patients as described previously.9 To explore their influence on antibiotic resistance, differences between patients infected with the two most commonly shared Australian strains (AUST-01 and AUST-02) and those with unique strains were examined. To increase the likelihood that antibiotic susceptibility profiles from local laboratories reflected strains genotyped by the reference laboratory, this analysis was limited to patients with the same ERIC-PCR genotype for each of the three isolates typed.
Between-state comparisons
Health-care in Australia is administered at a state level. Consequently, to determine whether the state in which care was delivered influenced treatment and antibiotic resistance, between-state differences were compared. Each of the analysed states is identified by the letters A to F, with the suffix (p) or (a) denoting paediatric and adult centres, respectively.
Statistical analysis
Data analysis was performed using PASW, Version 18.0 (SPSS Inc., Chicago, IL, USA). Between-group differences of continuous variables were analysed by t-tests, analysis of covariance, Mann–Whitney or Kruskal–Wallis tests according to the normality of the data distribution. Correlations were examined by Spearman's Rho and categorical variables by Chi-squared tests. Multi-logistic and linear regression analyses were performed to identify independent predictors of MARPA and intravenous antibiotic usage in the previous 12 months, respectively. To determine the impact of shared strains, regression models were rerun on a subgroup of 574 subjects harbouring either AUST-01, AUST-02 or a unique P. aeruginosa strain, with the variable P. aeruginosa strain type incorporated. In all models, State A (coordinating State for ACPinCF) acted as the reference state, and variables were included in each model if they achieved P-values <0.05 in their respective univariate analysis. A P-value <0.05 was considered statistically significant.
Results
Antibiotic resistance data were available for at least one isolate from 934/983 (95%) patients (228 children) culturing P. aeruginosa from their sputum. Forty-eight per cent of patients were p.Phe508del homozygotes, 32% p.Phe508del heterozygotes, 3% had other mutations and 17% had unknown CF genotypes. The clinical characteristics and isolate antibiotic resistance profiles of paediatric and adult patients treated in each state are summarized in Tables 1 and 2 and for individual treatment centres included in the study in Supplementary Tables S2 and S3, respectively.
|
Whole cohort (n = 228) |
State A (n = 81) |
State B (n = 62) |
State C (n = 62) |
State D (n = 23) |
P-valuec | |
|---|---|---|---|---|---|---|
| Age | 14 (11–17) | 14 (11–16) | 14 (11–16) | 14 (9–17) | 17 (15–19) | 0.001 |
| Male, number (%) | 114 (50) | 34 (42) | 30 (48) | 34 (55) | 16 (70) | 0.10 |
| FEV1 % predicteda | 74 (55–88) | 77 (55–90) | 74 (61–87) | 68 (53–87) | 73 (46–87) | 0.83 |
| FVC % predicteda | 88 (70–98) | 89 (67–99) | 87 (75–98) | 87 (68–97) | 88 (71–96) | 0.90 |
| BMI z-scorea | −0.35 (−0.96–0.23) | −0.21 (−0.61–0.27) | −0.62 (−1.08–−0.15) | −0.32 (−1.00–0.36) | −0.02 (−1.32–0.28) | 0.30 |
| No. of outpatient visitsa | 6 (4–8) | 6 (4–9) | 5 (2–7) | 6 (4–7) | 7 (5–7) | 0.58 |
| No. of IV antibiotic coursesa | 1 (0–3) | 1 (0–3) | 0 (0–2) | 1 (0–2) | 3 (1–4) | <0.001 |
| No of IV antibiotic daysa | 14 (0–35) | 20 (1–41) | 0 (0–21) | 10 (0–23) | 55 (19–81) | <0.001 |
| % Inhaled antibioticb | 46 | 30 | 53 | 54 | 67 | 0.002 |
| % Dornase-alphab | 56 | 57 | 57 | 46 | 81 | 0.06 |
| % HTSa | 29 | 33 | 24 | 9 | 77 | <0.001 |
| % Azithromycinb | 54 | 62 | 52 | 37 | 73 | <0.01 |
| P. aeruginosa antibiotic resistance, % | ||||||
| - Aminoglycosides | 43 | 56 | 36 | 24 | 74 | <0.001 |
| - β-lactam agents | 21 | 31 | 8 | 5 | 61 | <0.001 |
| - Fluoroquinolones¶ | 30 | 56 | 24 | 16 | 39 | <0.001 |
| - MARPA†† | 31 | 51 | 17 | 10 | 65 | <0.001 |
- Due to the skewed distributions of the data the differences between continuous variables were examined by the Kruskal–Wallis test and reported as medians and interquartile ranges. Differences between categorical variables were examined by chi-squared tests. P-value <0.05 considered statistically significant.
- a In the 12 months prior to sample collection, for FEV1% predicted, this was the best value recorded in the calendar year and the FVC and BMI recorded on that day.
- b In the 30 days prior to sample collection.
- c Variance between treatment centres.
- Antibiotic susceptibility testing data unavailable on ¶14 samples, ††12 samples.
- BMI, body mass index; FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; HTS, hypertonic saline; MARPA, multi-antibiotic resistant P. aeruginosa.
| Whole cohort (n = 706) |
State-A (n = 222) |
State-B (n = 128) |
State-C (n = 174) |
State-D (n = 72) |
State-E (n = 92) |
State-F (n = 18) |
P-value§ | |
|---|---|---|---|---|---|---|---|---|
| Age | 27 (22–33) | 26 (22–33) | 27 (23–35) | 29 (24–38) | 27 (23–32) | 27 (23–35) | 27 (23–35) | 0.013 |
| Male, number (%) | 409 (58) | 101 (59) | 63 (58) | 70 (57) | 49 (68) | 48 (52) | 12 (67) | 0.42 |
| FEV1 % predicted† | 53 (38–69) | 51 (36–68) | 59 (41–74) | 47 (34–63) | 57 (37–78) | 49 (36–64) | 49 (36–64) | 0.026 |
| FVC % predicted† | 73 (58–86) | 73 (55–87) | 76 (61–94) | 68 (57–83) | 75 (60–91) | 65 (52–77) | 65 (52–77) | <0.001 |
| BMI† | 21 (19–24) | 21 (19–25) | 22 (20–25) | 21 (20–24) | 22 (20–23) | 22 (19–25) | 22 (19–25) | 0.26 |
| No. of outpatient visits† | 5 (3–9) | 7 (4–10) | 3 (1–6) | 6 (5–10) | 3 (1–5) | 6 (2–11) | 8 (4–12) | <0.001 |
| No. of IV antibiotic courses† | 1 (0–2) | 1 (0–3) | 0 (0–1) | 2 (1–3) | 1 (0–3) | 1 (0–2) | 1 (0–2) | <0.001 |
| No of IV antibiotic days† | 14 (0–32) | 15 (0–36) | 0 (0–14) | 21 (7–34) | 14 (0–38) | 14 (0–32) | 7 (0–22) | <0.001 |
| % Inhaled antibiotica | 48 | 32 | 57 | 61 | 38 | 52 | 71 | <0.001 |
| % Dornase-alphaa | 50 | 39 | 48 | 60 | 61 | 44 | 94 | <0.001 |
| % HTS† | 32 | 26 | 33 | 39 | 33 | 27 | 53 | <0.05 |
| % Azithromycina | 65 | 57 | 84 | 72 | 52 | 55 | 59 | <0.001 |
| P. aeruginosa antibiotic resistance, % | ||||||||
| - Aminoglycosides¶ | 53 | 44 | 63 | 60 | 60 | 45 | 50 | 0.001 |
| - β-lactam agents†† | 16 | 21 | 18 | 16 | 6 | 10 | 6 | 0.02 |
| - Fluoroquinolonesb | 46 | 35 | 38 | 75 | 35 | 37 | 67 | <0.001 |
| - MARPAb | 35 | 25 | 40 | 55 | 21 | 24 | 44 | <0.001 |
- Due to the skewed distributions of the data the differences between continuous variables were examined by the Kruskal–Wallis test and reported as medians and interquartile ranges. Differences between categorical variables were examined by chi-squared tests. P-value <0.05 considered statistically significant †In the 12 months prior to sample collection, for FEV1% predicted this was the best value recorded in the calendar year and the FVC and BMI recorded on that day.
- a In the 30 days prior to sample collection.
- §Variance between treatment centres. Antibiotic susceptibility testing data unavailable on ¶one sample, ††two samples,
- b seven samples.
- BMI, body mass index; FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity; HTS, hypertonic saline; MARPA, multi-antibiotic resistant P. aeruginosa.
Paediatric centres
Eight paediatric centres (4–62 patients contributing P. aeruginosa isolates) from four states provided samples. Of the infected patients, 31% had MARPA isolates, with 43%, 21% and 30% of all isolates showing resistance to aminoglycoside, β-lactam and fluoroquinolone antibiotics, respectively (Table 1, Supplementary Table S2). MARPA and resistance to β-lactam and fluoroquinolone antibiotics correlated with an increased number of intravenous antibiotic days and outpatient visits in the preceding year (Supplementary Table S4).
Except for higher age in state-D (p), patient characteristics were similar between states (Table 1). In contrast, large differences in intravenous antibiotic use, maintenance therapies and antibiotic resistance rates existed between states. Logistic regression demonstrated the state where treatment was delivered was independently associated with MARPA, while linear regression found azithromycin, reduced FEV1% predicted and state were independently associated with intravenous antibiotic usage (Tables 3 and 4, respectively).
| Variable | Paediatric centres | Adult centres | ||
|---|---|---|---|---|
| Odds ratio (95% CI) | P-value | Odds ratio (95% CI) | P-value | |
| Univariate | ||||
| Age | 1.03 (0.97–1.11) | 0.317 | 1.01 (0.98–1.02) | 0.767 |
| Female gender | 0.83 (0.46–1.49) | 0.531 | 0.99 (0.72–1.35) | 0.940 |
| FEV1% predicted† | 0.56 (0.14–2.24) | 0.416 | 0.57 (0.27–1.22) | 0.151 |
| BMI‡ | 1.14 (0.82–1.57) | 0.436 | 0.94 (0.90–0.98) | 0.009 |
| Inhaled antibiotic§ | 0.60 (0.32–1.11) | 0.104 | 1.33 (0.97–1.82) | 0.080 |
| Dornase-alpha§ | 1.71 (0.93–3.16) | 0.085 | 1.46 (1.07–2.00) | 0.017 |
| HTS‡ | 2.09 (1.12–3.92) | 0.021 | 1.49 (1.07–2.07) | 0.018 |
| Azithromycin§ | 1.76 (0.96–3.21) | 0.065 | 1.51 (1.08–2.11) | 0.017 |
| No. of IV antibiotic days¶ | 1.01 (1.00–1.02) | 0.001 | 1.00 (1.00–1.01) | 0.421 |
| States A | 1.00 (ref) | 1.00 (ref) | ||
| B | 0.20 (0.09–0.47) | <0.001 | 1.95 (1.22–3.11) | 0.005 |
| C | 0.11 (0.04–0.29) | <0.001 | 3.63 (2.37–5.58) | <0.001 |
| D | 1.93 (0.72–5.14) | 0.188 | 0.77 (0.41–1.48) | 0.439 |
| E | — | 0.93 (0.52–1.63) | 0.790 | |
| F | — | 2.65 (1.02–36.86) | 0.045 | |
| Multivariate | ||||
| BMI | †† | 0.9 (0.89–0.99) | 0.012 | |
| State A | 1.00 (ref) | 1.00 (ref) | ||
| B | 0.20 (0.09–0.47) | <0.001 | 2.06 (1.28–3.30) | 0.003 |
| C | 0.11 (0.04–0.29) | <0.001 | 3.77 (2.44–5.82) | <0.001 |
| D | 1.93 (0.72–5.14) | 0.188 | 0.80 (0.42–1.53) | 0.496 |
| E | — | 1.00 (0.56–1.77) | 0.989 | |
| F | — | 2.17 (0.77–6.14) | 0.142 | |
- †Best FEV1 recorded in the calendar year. ‡Z-scores used for paediatric data.
- §In the 30 days prior to sample collection. ¶In the 12 months prior to sample collection. ††Not significant in the final multivariate model.
- BMI, body mass index; FEV1, forced expiratory volume percentage; HTS, hypertonic saline; IV, intravenous.
| Variable | Paediatric centres | Adult centres | ||
|---|---|---|---|---|
| β (95% CI) | P-value | β (95% CI) | P-value | |
| Univariate | ||||
| Age | 1.25 (0.19–2.31) | 0.021 | −0.03 (−0.28–0.23) | 0.839 |
| Female | 7.24 (−1.87–16.35) | 0.119 | 9.28 (4.87–13.70) | <0.001 |
| FEV1% predicteda | −80.70 (−99.55–−61.85) | <0.001 | −51.89 (−61.83–−41.95) | <0.001 |
| BMIb | −7.17 (−12.10–−2.24) | 0.005 | −1.96 (−2.57–−1.35) | <0.001 |
| Inhaled antibioticc | 15.61 (6.27–24.95) | 0.001 | 7.21 (2.71–11.71) | 0.002 |
| Dornase-alphac | 11.82 (2.65–21.00) | 0.012 | 10.35 (6.00–14.69) | <0.001 |
| HTSb | 22.02 (12.09–31.94) | <0.001 | 6.95 (2.24–11.66) | 0.004 |
| Azithromycinc | 26.07 (17.38–34.76) | <0.001 | 12.62 (8.07–17.17) | <0.001 |
| States A | 1.00 (ref) | 1.00 (ref) | ||
| B | −13.93 (−24.93–−2.933) | 0.013 | −17.78 (−24.13–−11.44) | <0.001 |
| C | −13.23 (−24.22–−2.33) | 0.019 | −1.47 (−7.29–4.33) | 0.618 |
| D | 26.04 (11.04–41.04) | 0.001 | −1.49 (−9.26–6.27) | 0.706 |
| E | — | −6.37 (−13.44–0.70) | 0.077 | |
| F | — | −12.55 (−26.86–1.76) | 0.086 | |
| Multivariate | ||||
| Female gender | d | 5.75 (1.60–9.90) | 0.007 | |
| BMI | d | −0.80 (−1.40–−0.20) | 0.009 | |
| Dornase alphac | d | 4.91 (0.69–9.14) | 0.023 | |
| Azithromycinc | 12.91 (4.71–21.10) | 0.002 | 8.17 (3.67–12.67) | <0.001 |
| FEV1% predicted | −71.03 (−88.50–−53.57) | <0.001 | −39.21 (−49.55–−28.87) | <0.001 |
| State A | 1.00 (ref) | 1.00 (ref) | ||
| B | −14.01 (−23.61–−4.41) | 0.004 | −17.63 (−3.56–−11.70) | <0.001 |
| C | −13.54 (−24.29–−2.79) | 0.014 | −3.52 (−8.94–1.89) | 0.202 |
| D | 23.33 (9.69–36.99) | 0.001 | 0.64 (−6.59–7.87) | 0.862 |
| E | — | −6.69 (−13.16–−0.22) | 0.043 | |
| F | — | −14.24 (−27.82–−0.66) | 0.040 | |
- a Best FEV1 recorded in the calendar year.
- b Z-scores used for paediatric data.
- c In the 30-days prior to sample collection.
- d Not significant in the final multivariate model.
- BMI, body mass index; FEV1%, forced expiratory volume percentage; HTS, hypertonic saline.
Adult centres
Ten adult centres (15–171 patients contributing P. aeruginosa isolates) from six states participated. Of the infected patients, 35% had MARPA, with 53%, 16% and 46% showing resistance to aminoglycoside, β-lactam and fluoroquinolone antibiotics, respectively (Table 2, Supplementary Table S3). In contrast with paediatric centres, the number of intravenous antibiotic courses and days did not correlate with antibiotic resistance. However, a weak correlation existed between number of outpatient visits, maintenance therapies (mucolytics and azithromycin) and MARPA isolates (Supplementary Table S4).
Comparisons of states demonstrated greater heterogeneity across clinical parameters than seen in paediatric states. Similarly, there were significant differences in intravenous antibiotic and maintenance treatment usage, and rates of antibiotic resistance between adult centres (Table 2). BMI and state where treatment was received were independent determinants of MARPA (Table 3). Being female, lower BMI, recent use of alpha dornase and azithromycin, reduced FEV1% predicted and state were independently associated with intravenous antibiotic usage (Table 4).
Shared P. aeruginosa strains
Of the 934 patients, 574 (435 adult, 139 paediatric) had the same unique (289), AUST-01 (160) or AUST-02 (125) genotype for each of their three isolates typed. Among this subgroup, rates of AUST-01 (40% vs 22%, P < 0.001) and AUST-02 (34% vs 20%, P < 0.01) were each significantly higher for patients in adult, compared with paediatric centres. Clinical parameters between the three groups were similar, however, compared with patients with unique isolates, AUST-01 and AUST-02 infected patients were more likely to have MARPA, aminoglycoside or β-lactam resistant isolates (Supplementary Table S5). Moreover, AUST-01 isolates had higher fluoroquinolone resistance rates. Among adults, infection with a shared strain was an independent predictor of increased intravenous antibiotic usage (Supplementary Tables S6 and S7).
Testing methodology and MARPA
Overall, 743 (79.6%) sputum samples were collected during outpatient visits. Within each state, MARPA prevalence did not differ significantly between inpatient and outpatient collected samples (Supplementary Table S8) or between mixed and pure isolate selection techniques (30.4% vs 35.4%, respectively, P = 0.16; Supplementary Table S9). Although samples analysed using CDS or CLSI breakpoints provided similar MARPA rates (28.6% vs 30.1%, P = 0.76), these were significantly higher (62.3%) in the single centre (C(a)2) employing EUCAST breakpoint values (Supplementary Table S9).
Discussion
By linking data from the ACFDR and the national ACPinCF study, we found that Australian CF patients harbouring P. aeruginosa had MARPA rates, intravenous antibiotic usage and maintenance treatment practices that varied noticeably between states. In children, a state in which care was delivered was the only measured factor to be independently associated with MARPA, while increased intravenous antibiotic treatments were associated with recent use of azithromycin, reduced lung function and treating state. In adults, in addition to treating state, BMI was associated with MARPA, while being female, BMI, recent inhaled dornase-alpha, use of azithromycin, reduced lung function and treating state were independently associated with intravenous antibiotic usage.
Inhaled tobramycin, frequent intravenous antibiotics, CF-related diabetes, hospitalization for acute exacerbations and treatment in a centre with an already high MARPA prevalence were identified previously to be associated with MARPA in CF patients with newly acquired P. aeruginosa infection.20 Longitudinal studies of P. aeruginosa infection in non-CF individuals have also found MARPA prevalence closely parallels intravenous antibiotic use.21, 22 Thus, our observation that MARPA in CF adults did not correlate with intravenous antibiotic usage or strongly with maintenance treatment burden was unexpected. This may be explained partly by the current study examining prevalence of MARPA, as opposed to the incidence of new MARPA cases. Consequently, we speculate that in some patients, infection may have been longstanding in which a complex array of factors might have influenced P. aeruginosa phenotype over time. This would include a greater impact from shared strains, phenotypic adaptation following cumulative antibiotic exposure, host immune responses and the local CF lung environment.23-26 For example, despite similar lung function and clinicians being unaware of strain typing status, patients with highly prevalent AUST-01 and AUST-02 strains were more likely to receive greater intensity of antibiotic treatments than those infected with unique strains. Furthermore, we cannot exclude the possibility that the treating clinician's knowledge and attitudes towards MARPA infection may have influenced their treatment approach within their centre.
It is important to highlight that in chronic P. aeruginosa infection, antibiotic susceptibility does not predict clinical response to treatment, and the clinical significance of MARPA in CF remains contentious.27, 28 While a single-centre study involving CF adults reported an increased rate of pulmonary decline in patients with MARPA infection,29 this was not supported by a recent large data registry-based study, which found instead MARPA were more likely markers for advanced disease or intensive antibiotic therapy.30 Furthermore, a CF Foundation Benchmarking Project identified early and aggressive antibiotic management of pulmonary decline to be a distinguishing feature of CF centres (paediatric and adult) achieving outstanding clinical outcomes.8
The prevalence of MARPA among paediatric centres is comparable with rates reported in an earlier single-centre study.31 Indeed, despite higher antibiotic resistance, there was no association between MARPA and adverse clinical outcomes. Nevertheless, concerns remain over increasing treatment and its complexity in response to MARPA infection, which could lead to treatment-related morbidity and emergence of shared strains.32-35
These findings underline the dilemma facing CF physicians. Maintaining lung function and quality of life, while attempting to minimize antibiotic resistance and treatment-related adverse effects is especially challenging when the mechanisms of antibiotic action in CF lung disease are incompletely understood.36 In addition to developing new antibiotics and delivery devices, further study is needed urgently to identify meaningful minimum inhibitory concentration (MIC) clinical breakpoints for respiratory bacterial pathogens within the CF lung and reliable non-invasive biomarkers to guide antibiotic management of acute pulmonary exacerbations.
Our study has several important limitations. First, in vitro antibiotic susceptibility testing methods varied between CF centres (Supplementary Table S1). However, there are no standardized approaches to antibiotic susceptibility testing nationally or internationally.37 For logistical and funding reasons, it was not possible to employ a standard method to confirm the susceptibility of each isolate submitted to the research laboratory. However, antibiotic resistance rates did not differ significantly between mixed versus pure testing methodologies. Furthermore, despite CLSI applying higher MIC breakpoints than CDS for determining resistance to β-lactam antibiotics (Supplementary Table S10), centres applying CLSI methodology generally showed higher rates of antibiotic resistance (Supplementary Tables S2 and S3). These data indicate that differences between CDS and CLSI testing methodologies are unlikely to have contributed to our findings. MARPA rates were however higher in the single centre using EUCAST MIC breakpoints, but we were unable to determine if this was from testing methodology or genuinely higher resistance rates. Second, our cohort included only those capable of providing sputum. This meant that patients with more severe lung disease were likely to be selected, as younger patients and those with mild disease often cannot produce a sputum sample on demand.38, 39 Third, as this study was cross-sectional, we could not differentiate between those with chronic or intermittent infections. This is especially relevant in young children where newly infecting strains are more likely to be susceptible to anti-pseudomonal antibiotics. However, paediatric participants were on average 6 years older than the mean age of the clinic population from where they were selected (data not shown) and had relatively high rates of antibiotic resistance, suggesting that most were chronically infected. Fourth, the cross-sectional design also meant that we could not ascertain if increased antibiotic resistance in adults was ‘imported’ from a paediatric centre with high antibiotic use or resulted from a lifetime of accumulated antibiotic exposure, and whether MARPA isolates were associated directly with worse clinical outcomes. Fifth, we did not collect information on the type of intravenous antibiotics prescribed for each treatment course, which may have varied between centres and influenced induction of antibiotic-resistant strains.35 Finally, maintenance therapies were recorded from hospital dispensing data systems and patient medical records, which may not represent actual treatment adherence.
In conclusion, our national study found large between-centre differences in treatment practices and P. aeruginosa antibiotic susceptibility profiles. Increased antibiotic resistance in paediatric, but not adult, centres was associated with intravenous antibiotic usage, while in both paediatric and adult centres MARPA was not associated with worse lung function. We plan to continue monitoring this cohort longitudinally to determine whether MARPA isolates do actually result in worse clinical outcomes or simply reflect more advanced disease and/or intensive treatment strategies adopted by some centres. This study emphasizes the need to develop a greater mechanistic understanding of antibiotic actions in CF lung disease so that they can be used most effectively and safely.
Acknowledgements
We are grateful to all the participants, CF Centre Directors, research and clinical staff and clinical microbiology laboratories at each of the study sites. We also thank Geoff Sims (Cystic Fibrosis Australia) for his assistance with the ACFDR data. For participating laboratories and co-investigators of the ACPinCF study see Appendix in the Supplementary Information to this publication. This work was supported by the National Health and Medical Research Council Project Grant 455919, The Children's Hospital Foundation, Queensland Health Office of Health and Medical Research, the Australian Cystic Fibrosis Research Trust and The Prince Charles Hospital Foundation and Rotary Australia. The funding sources had no role in the design and conduct of the study, analysis or interpretation of the data, preparation or final approval of the manuscript or the decision to submit the manuscript for publication.
