Reduction in Pseudomonas aeruginosa sputum density during a cystic fibrosis pulmonary exacerbation does not predict clinical response

Background Pulmonary exacerbations (PEx) are critical events in cystic fibrosis (CF), responsible for reduced quality of life and permanent loss of lung function. Approximately 1/4 of PEx are associated with failure to recover lung function and/or resolve symptoms. Developing tools to optimize PEx treatment is of paramount importance. Methods We retrospectively audited all adults infected with Pseudomonas aeruginosa, experiencing PEx necessitating parenteral antibiotic therapy from 2006–2012 from our center. Quantitative analysis of sputum at admission, twice-weekly during hospitalization, and end of therapy were compared to baseline (most recent healthy) and follow-up (after PEx) samples. Change in P. aeruginosa burden from baseline was assessed for any and all morphotypes (ALL), as well as mucoid (MUC) and non-mucoid (NON) isolates specifically. PEx were identified as failures if >90% of baseline pulmonary function was not recovered. Results Forty-six patients meeting the above inclusion and exclusion criteria experienced 144 PEx during this time (median 3, IQR 2–6). Patients were treated for a median 14 days (IQR 13–16). No increase in ALL, MUC or NON were detected at PEx, nor was there an association between change in sputum density and magnitude of lung function decline. PEx failures were observed in 30% of events. Reductions of at least 1-log and 2 log P. aeruginosa sputum density was observed in 57% and 46% (ALL), 73% and 55% (MUC) and 58% and 46% (NON) of PEx, respectively. Factors associated with greater reduction of P. aeruginosa sputum density included choice of β-lactam antibiotic, antibiotics with in vitro predicted activity and treatment duration. PEx associated with reductions in P. aeruginosa sputum density were not associated with a reduced risk of PEx failure. Conclusions Enhanced killing of P. aeruginosa during PEx does not predict improved clinical outcomes. Studies accounting for the polymicrobial nature of CF respiratory disease and the heterogeneity of P. aeruginosa causing chronic infection may enable the identification of a more appropriate pathogen(s) based biomarker of PEx outcomes.


Background
The archetypal CF pathogen, Pseudomonas aeruginosa infects 50-70% of patients [1]. Patients with chronic P. aeruginosa infection have increased rates of lung function decline, health care utilization, and reduced survival [2][3][4]. Chronic P. aeruginosa infection is punctuated by frequent acute deteriorations termed pulmonary exacerbations (PEx). PEx are characterized by increased cough and sputum production, disproportionate shortness of breath and loss of lung function, as well as increased inflammation [5][6][7]. PEx are critical events in CF, associated with reduced quality of life [8], increased cost [9], permanent lung damage [10,11] and increased short-term mortality [12,13]. So important are these events, they may now constitute primary end-points in CF therapeutic trials [14].
Treatment of PEx usually consists of aggressive airway clearance, respite, nutritional support, and antimicrobial therapy directed against chronically infecting pathogens. Despite therapy, 25% of PEx fail to achieve successful outcomes as determined by lung function recovery, resolution of symptoms and preventing recurrences [15]. Patients more likely to experience unsuccessful PEx outcomes are infected with MRSA, Burkholderia cepacia complex, MDR (multi-drug resistant) P. aeruginosa, and those with CF-related diabetes and/or CF-liver disease [10,15]. Furthermore, more severe PEx including those with greater initial lung function decline, and those with greater inflammation at initiation and completion of treatment are associated with worse outcomes.
Developing tools to predict PEx outcomes is key to their successful management. It has been previously determined that antibiogram directed therapy against a limited number of isolates of P. aeruginosa has only a weak association with PEx outcomes [15]. As such, other biomarkers are increasingly being evaluated for their ability to predict treatment responses. A number of host specific factors are being assessed [16,17]. However, given the critical intervention in PEx is anti-bacterial, the use of bacterial derived biomarkers to follow treatment response deserves attention. While antibacterials have been shown to reduce the bacterial load during the treatment of PEx, how this correlates with clinical response has not been established [18]. Herein we evaluate the use of semi-quantitative reporting of P. aeruginosa sputum density and correlated the response with clinical outcomes during PEx treatment.

Methods
All CF patients chronically infected with P. aeruginosa attending the Calgary Adult CF Clinic from 2006-2012 experiencing PEx treated with parenteral antibiotics were considered for inclusion if they had semi-quantitative sputum cultures performed ≥3 times during treatment (baseline, initiation, early, end of therapy) and at followup. Parenteral antibiotics provided for reasons other than PEx were excluded. Patients were excluded if they had a baseline FEV 1 < 30% predicted, were infected with Burkholderia cenocepacia or Mycobacterium abscessus, or were listed for lung transplantation. Detailed review of clinical records were performed from prior to the PEx, through treatment and in follow-up. Pulmonary function was evaluated by spirometry. Data was prospectively collected and retrospectively analyzed. Patients provide prospective consent for studies correlating clinical outcomes with lower airways infection. This project has been approved by the Conjoint Health Research Ethics Board of the University of Calgary (E23087).

Definitions
Baseline lung function was defined as the best forced expiratory volume in one second (FEV 1 ) percent predicted in the preceding 6 months. The Leeds criteria was used to define chronic P. aeruginosa infection [19]. PEx were diagnosed on clinical grounds in real time, however, charts were evaluated and scored according to Fuchs criteria retrospectively [5]. Parenteral therapies were provided as either an inpatient, outpatient via the home parenteral therapy program (HPTP) or a combination thereof. Treatment failures were defined a priori as failure to have recovered 90% of baseline FEV 1 at follow-up (FTR90) [10,11,15]. Severity of lung disease was categorized as mild (FEV 1 % predicted ≥70%), moderate (40-70%), and severe (<40%). Bacterial sputum density is reported as log10 CFU/ml. Bacteriologic responses to antipseudomonal therapies were evaluated as either a one or two-log drop in CFU/ml.

Microbiologic investigations
Baseline sputum samples were from the most recent clinic where patients were not experiencing PEx. Twiceweekly sputum samples were collected through treatment. Admission samples were collected before starting therapy. Changes in bacterial load were evaluated at "EARLY" (3 +/− 2 days from starting) and "END" (within 3 days of completion). Follow-up samples were collected at the next clinic visit attended.
Sputum samples were mixed with equal volumes of dithiothreitol (DTT, Sputolysin, EMD Millipore, MA) and vortexed. Samples were serial diluted in DTT to 10 −5 and plated on Columbia Blood Agar (CBA), MacConkey Agar, (MAC), Chocolate Agar (CHOC), Burkholderia cepacia selective agar, and Mannitol Salt Agar. Pathogens were identified by standard criteria [20]. Oral flora (OF) included organisms such as non-hemolytic Streptococci, Haemophilus parainfluenzae, Corynebacterium spp., Neisseria spp. (non-meningitidis) and Coagulase-negative Staphylococci grown on CBA or CHOC. Organisms were quantified and P. aeruginosa colonies classified as mucoid (MUC) or non-mucoid (NON) isolates based on growth on MAC [21]. P. aeruginosa susceptibility testing was performed on each morphotype by Kirby-Bauer or E-test as per CLSI guidelines [22]. Where >1 MUC/NON isolates existed susceptibility testing was performed on each, and the result represents the least susceptible. MDR was defined as resistance to all antibiotics in ≥2 classes and pan-drug resistance (PDR) all classes, save for colistin [15,23].

Statistical analysis
Means with standard deviation (SD) or medians with inter-quartile ranges (IQR) were used to describe normal and non-normally distributed data, respectively. Categorical variables were compared using a two-sided Fisher's exact test. Odds ratios (OR) were calculated by dividing the proportion with a factor to those without and were reported with 95% confidence intervals (CI). Continuous variables were compared using Pearsons correlation coefficient. Statistical analyses were performed using Stata 12.0 (Stata Corp., College Station, TX, USA). A p value of ≤0.05 was considered significant.

Patients and PEx
Of a total eligible cohort of 176, 46 patients met inclusion criteria. These 46 patients experienced 144 PEx during the six-year study period. Median age of those included was 32.9 years (IQR 27.9-40). Patients were diagnosed with CF at a median age of 1.35 years (IQR 0.5-6). 56% were F508del homozygous and 83% carried ≥1 allele. 43 patients (93%) were pancreatic insufficient. Co-morbidities included CF-related diabetes in 18 (40%), CF-liver disease 8 (18%), osteopenia 60% and osteoporosis 9%. During the six years of study, 14 patients initially included had future events excluded because of progression of disease and three patients died, although none from CF.
Patients presented with an exacerbation 68 days (IQR 32-111) after their most recent clinic visit. Mean duration from initiation of antibiotics to collection of EARLY sputum culture was 3.84 days (SD 1.75). Mean duration from the end of therapy to collection of END sputum sample was 0.72 days (SD 1.70). 40 days (IQR 27-61) after completion of PEx treatment, patients were seen in follow-up at a regular CF clinic.

PEx outcomes
PEx treatment failures, as defined by failing to recover 90% of baseline FEV 1 , were observed in 42/140 (30%). Neither sign, symptom, nor total Fuchs' initial PEx diagnostic score were associated with increased failure risk (not shown). Stage of CF-lung disease did not influence failure (not shown). Antibiotic courses ≤10 days were more likely to be deemed failures; 8/16 (50%) vs 38/128 (30%), OR 1.68 (0.97-2.94), p = 0.15) but this did not reach statistical significance. Few prior and concomitant medications influenced PEx outcomes. In particular, prior tobramycin inhalation solution was associated with a trend to reduced risk of failure [
Clinical outcomes were not predicted based on changes in sputum bacterial burden at completion of PEx treatment (Table 1) Figures 2  and 3). Relative to other antibiotics ceftazidime showed a modest advantage, whereas cefepime was associated with reduced P. aeruginosa killing. Tobramycin dosing once or twice daily did not influence bacteriologic response (data not shown). An association between number of antibiotics with in vitro predicted activity and reduction of P. aeruginosa density was observed although this reached statistical significance only for regimens with activity against the NON-isolates (Table 3).
No significant impact of concurrent therapies was noted on reductions in P. aeruginosa sputum density. In particular, concurrent chronic azithromycin, DNase, nor hypertonic saline impacted P. aeruginosa density. A trend was observed in individuals receiving concurrent ICS at having a lower likelihood of reducing P. aeruginosa sputum burden (not shown). Venue for antibiotic delivery did not impact bacteriologic response (not shown). Treatments <14 days produced an inferior bacteriologic response, although this met statistical significance for only the NON-isolates (Table 2).
MSSA was a chronic colonizer of patients in 31/146 (21.2%) of PEx, and two had a history of recent-prior MRSA transient colonization, though none on admission.  (Figure 4). EARLY changes in bacterial load did not, however, predict failure risk for any P. aeruginosa morphotype (not shown).

Discussion
PEx are increasingly recognized as a significant contributor to the progressive nature of CF lung disease [24]. Avoiding PEx is critical in this regard, and many CF therapies may exert their beneficial effects in manner [5,[25][26][27]. However, despite attempts to limit their occurrence, PEx remain common. Given that some PEx are associated with worse outcomes than others, strategies to identify and optimize the management of these events are critical.
Anti-pseudomonal antibiotics are an integral component of PEx management. Independent of enhanced airway clearance, anti-pseudomonal antibacterials have been shown to improve pulmonary function in a linear capacity relative to reducing P. aeruginosa burden [18]. Few studies however, have tried to identify factors associated with enhanced P. aeruginosa killing. Smith et al., followed 75 patients chronically infected with P. aeruginosa through PEx managed with single or dual agent anti-pseudomonals [28]. While they observed a general reduction in sputum density of P. aeruginosa, reduced sputum DNA and protein levels, none of these factors correlated with absolute change in pulmonary function. Similarly, Mclaughlin et al., did not observe a correlation between burden of P. aeruginosa and absolute lung function recovery [29]. More recently, Deschaght et al., performed an observational study where they identified a correlation between improvement in lung function and reduction in P. aeruginosa density using quantitative PCR [30]. These studies, however, focused on aggregate analysis of lung function improvement and did not categorize individual events on the basis of success or failure, nor assess impact of individual morphotypes.
We sought to determine if bacteriologic killing as a measure of effectiveness of antibacterial therapies correlated with clinical outcomes in PEx. During antibacterial treatment of PEx, a progressive decline in P. aeruginosa burden was noted in approximately 60% of patients. This decline was predictable based on early changes in P. aeruginosa burden after only four days of treatment. Reduction in total P. aeruginosa burden, nor mucoid or nonmucoid isolates specifically did not, however, correlate with risk of failing to recover lung function. Enhanced  killing of P. aeruginosa was observed in patients receiving antibacterials estimated to have more potent in vitro activity. This was particularly true when antibiotics had greater activity against non-mucoid isolates that were more readily reduced with antibiotics. Furthermore, we have observed like others, that in those few patients who transiently achieve clearance of P. aeruginosa during PEx treatments, no enhancement in clinical outcomes was afforded [29].
To complicate matters, studies have shown resolution of PEx in patients chronically infected with P. aeruginosa treated with regimens devoid of anti-pseudomonal activity suggesting response is more complicated than a one-host and one-pathogen model [31,32]. Accordingly, more recent studies have assessed the impact of antibacterials on other constituents of the CF respiratory microbiome, beyond classical CF pathogens. Using enhanced culturing to identify the Streptococcus milleri group (S. anginosus group, SMG), 40% of PEx were identified to have this organism emerge as numerically dominant at the onset of PEx, and resolution of symptoms correlated with reduction in sputum burden [31,33,34]. Using culture independent techniques, others have observed that with the exception of the Prevotella and Chrysiogenales, levels of  other microbiome constituents do not correlate with improvements in lung function [35,36]. We did not detect an increase in either total P. aeruginosa load, or specifically MUC or NON morphotypes associated with an exacerbation. Nor was there an association between change in P. aeruginosa sputum density and extent of decline in pulmonary function. This data mirrors that of other recent studies that have challenged the long held belief that an increase in P. aeruginosa triggers PEx [37][38][39]. Clearly, factors other than changes in macroscopically apparent P. aeruginosa morphotypes are involved. These factors may include: environmental exposures such as pollution [40] and allergens [41]; host factors such as medication compliance [42] and co-morbidities [15,43,44] pathogen factors such as respiratory viruses [45], changes in the constituents of the lower respiratory microbiome [33,46] and specific P. aeruginosa sub-populations [47,48]. Furthermore, it is likely that there are multiple convergent pathways leading to a PEx, and these events vary in their manifestations and potential outcomes.
Many contentious issues exist in PEx management strategies. Data in these areas are difficult to interpret as outcomes have traditionally focused on absolute improvement in pulmonary function, as opposed to a more appropriate model of proportional recovery [46,47] Whereas some, but not all studies have suggested HPTP to be inferior, we noted similar outcomes in risk of failures and reductions in P. aeruginosa burden [46,47]. Likewise length of therapy for PEx has been debated [49]. Herein we noted a trend towards increased risk of failure to recover lung function with therapies ≤10 days, but no benefit with extending therapies >14 days. We also observed treatments <14 days were inferior in reducing P. aeruginosa sputum density.
The effectiveness of different antibiotics in achieving good PEx outcomes has been previously assessed. Herein we did observe differential abilities of antibiotics to reduce P. aeruginosa burden. In particular, cefepime was inferior at reducing P. aeruginosa sputum burden, although no impact on PEx outcome was noted. Interestingly, metaanalyses have demonstrated patients treated with cefepime relative to an alternate β-lactam with a similar spectrum of activity may be associated with an increased risk of mortality [50,51]. While commonly used in CF, the only clinical studies evaluating cefepime in CF have been pharmacokinetic based [52,53]. Based on these data, further studies of cefepime usage in CF are advisable.
ICS are common agents in CF, with approximately one third of patients receiving them. However, their utility has not clearly been established and the general consensus is that they are over used and may even be harmful [54]. Indeed, in a randomized controlled trial to evaluate the effects of discontinuation of ICS in patients stably maintained on ICS showed equivalent outcomes [55]. ICS use has been associated with increased risk of PEx [43,56]. Our results now suggest an increased risk of PEx failure with ICS use, a trend towards reduced sputum density of P. aeruginosa during antibacterial treatment and may warrant further caution with these agents in CF.
Herein we have attempted to expand upon other studies assessing PEx outcomes and changes in P. aeruginosa sputum density. To do so we have utilized a proportional lung A B C Figure 4 Correlation in change in CFU from admission sputum sample Early in to exacerbation therapy (Y axis) to End of therapy (X axis). A). ALL P. aeruginosa morphotypes, B). Mucoid only isolates, C). Non-mucoid Isolates.
function recovery model to overcome the deficiency whereby other studies using absolute improvement fail to account for the extent of initial decline and therefore are poor predictors of success or failure [15]. While this work relies on clinician diagnosed PEx events, we have retrospectively confirmed they met diagnostic criteria using an established PEx diagnostic algorithm [5]. Furthermore, we have attempted to evaluate outcomes based on individual morphotypes of P. aeruginosa reported routinely by clinical laboratories [20]. Indeed, quantitative analysis of P. aeruginosa using qPCR would not distinguish between mucoid and non-mucoid isolates. As infection with mucoid isolates of P. aeruginosa is associated with reduced success of eradication treatment in patients with new infection [57,58] and a worse prognosis for patients with chronic lung infection [3,59], it might be expected that a differential response against mucoid isolates might impact PEx outcomes. This appears not to be the case.
Several limitations of this work deserve consideration. Because diagnosis of PEx were collected and analyzed retrospectively, the data is not as robust as prospectively collected data. Symptoms or signs not documented in clinical records may inadvertently be documented as absent. This may be somewhat mitigated, as these findings were collected at a single CF clinic over a short time period, where practice patterns were relatively uniform. Our study excluded PEx not requiring parenteral antibiotics. We observed a significantly higher rate of PEx failures than in other works [10,11,15]. While similar criteria were used in the selection of only parenteral antibiotic treated PEx, far fewer patients were treated with parenteral antibiotics in this study. This likely reflects institutional practice variation in the management of PEx and availability of local resources. Parenteral antibiotics were reserved for the most severe events as evidenced by the significant decline in lung function on admission. This is further supported by the relatively low level of P. aeruginosa drug resistance observed herein relative to other works [15,60]. Whether more aggressive management of exacerbations would result in improved outcomes is of interest.

Conclusions
In this work we identified that while the majority of patients treated with parenteral anti-pseudomonals experienced a temporary reduction in sputum burden, approximately 40% did not. This bacteriologic response could be predicted on the basis of sputum samples collected early into antibiotic treatment. Antibacterial treatments with a greater number of agents with in vitro predicted activity against all morphotypes, and in particular non-mucoid isolates was more likely to result in enhanced P. aeruginosa killing during PEx. However, patients who experienced enhanced killing of P. aeruginosa after completion of therapy were not afforded a reduced risk of PEx failure. Accordingly, bacteriologic response during PEx treatment cannot be used to predict success. Competing interests MDP has received research funding from Gilead sciences and participated in advisory boards for Gilead, Novartis and Roche. HRR has participated in advisory boards for Novartis and Vertex. JCL, RS and MGS report no conflicts.
Author contributions JCL was responsible for data collection, analysis and initial drafting of the manuscript. RS was responsible for data analysis, and manuscript revisions. MGS and HRR were responsible data maintenance, participated in study design and analysis, and assisted in manuscript preparation and revisions. MDP conceived the study, oversaw data collection and analysis and manuscript preparation and revisions. All authors read and approved the final manuscript.