The relationship of systemic inflammation to prior hospitalization in adult patients with cystic fibrosis

We enrolled 58 consecutive adult patients from the Cystic Fibrosis (CF) Clinic at St. Paul's Hospital (Vancouver, British Columbia, Canada) between April and December 2009, who were clinically stable at the time of assessment. For inclusion, patients had to have one or more clinical features consistent with the CF phenotype [12] as well as a genotype with two identifiable disease-causing CF transmembrane conductance regulator (CFTR) mutations and sweat chloride measurements greater than 60 mmol/L on two occasions. Patients who had an exacerbation within the previous 4 weeks were excluded from the study. This study was conducted with the approval of the University of British Columbia - Providence Health Care Research Ethics Board (UBC-PHC REB). Following informed consent, we collected venous blood samples and performed spirometry using standard techniques, in accordance with guidelines from the American Thoracic Society [13]. Demographic and clinical data were obtained by chart review.

Plasma was prepared from the collected blood samples and a select number of circulating inflammatory proteins were measured using high-sensitivity enzyme-linked immunosorbent assay (ELISA) kits that were commercially available. These included interleukin (IL)-6 and IL-1β (R&D Systems, Minneapolis, MN), cytokines involved in the early phase inflammatory response; C-reactive protein (CRP; R&D Systems), an acute phase response protein; granzyme B (GzmB; eBioscience, San Diego, CA), a protein involved in adaptive immunity; chemokine C-C motif ligand 18/pulmonary and activation-regulated chemokine (CCL18/PARC; R&D Systems) and surfactant protein D (SP-D; Biovendor, Brno, Czech Republic), pneumo-proteins (i.e. proteins synthesized predominantly in the lungs) [14, 15]; and LPS-binding protein (LBP; Hycult Biotech, Uden, The Netherlands) and soluble cluster of differentiation 14 (sCD14; R&D Systems), proteins involved in lipopolysaccharide (LPS) signalling. The coefficients of variation for these kits were 7.4%, 12.3%, 3.1%, 5.8%, 1.2%, 2.2%, 1.7%, and 5.9% respectively, and the lower detection limits were 0.039 pg/mL, 0.057 pg/mL, 0.010 ng/mL, 0.2 pg/mL, 0.01 ng/mL, and 0.2 ng/mL, 4.4 ng/mL, and 0.125 ng/mL respectively. LPS levels were measured using a commercially-available kinetic chromogenic Limulus amebocyte lysate (LAL) assay kit (Lonza Walkersville, Walkersville, MD) following plasma dilution and heat inactivation pre-treatment steps to diminish interference from plasma proteins. The coefficient of variation for the assay was 2.2%, and the lower detection limit was 0.5 pg/mL. Of these biomarker assays, only the IL-1β ELISA had samples with values below the limit of detection (18 undetectable out of 58 samples, or 31%). In plasma samples, the manufacturers' reported ranges for IL-6, IL-1β, CRP, GzmB, and sCD14 measurements in healthy volunteers were 0.428 to 8.8 pg/mL, non-detectable to 0.452 pg/mL, 104 to 4185 ng/mL, 0.8 to 24.1 pg/mL, and 1200 to 3100 ng/mL, respectively. Reference ranges were not available for CCL18/PARC, SP-D, LBP, and LPS, and the dynamic ranges of these assays were 7.8 to 500 pg/mL, 1.56 to 100 ng/mL, 4.4 to 50 ng/mL, and 0.5 to 500 pg/mL, respectively.

Data for the biomarker measurements were analyzed after natural log transformation due to their skewed distribution, and values below the limit of detection were assigned the value of the lower detection limit for the particular kit. The relationships of the biomarkers to lung function (FEV₁ as a percentage of predicted) and to each other were assessed using linear regression. Multivariable linear regression analysis was performed to assess the role of possible confounding factors such as age, sex, and pseudomonal status. Plasma biomarker levels were compared between those who did and did not experience a hospitalization in the past using a Student's t-test for independent samples. Fisher's exact test was used to analyze categorical data. The independent relationship of plasma biomarkers to FEV_1% predicted was ascertained using multiple regression analysis. To ensure parsimony and enhance the robustness of the model, we used a stepwise approach to select only those covariates that significantly impacted on the relationship (the significance level for entering, P ≤ 0.05 and the significance level for stay, P ≤ 0.05). To facilitate interpretation and cross-comparisons of beta-coefficients of plasma biomarkers from the regression model, we standardized the beta-coefficients to their standard deviation. Thus, the beta-coefficients are presented per 1 SD increase in the plasma biomarker expression. P-values less than 0.05 were considered significant (two-tailed test). All analyses were conducted using SAS (Carey, N.C.) version 9.1.

The age of the study subjects ranged from 18 to 61 years (Table 1). The baseline clinical characteristics of the study subjects are summarized in Table 1 in two groups based on the median value of FEV_1% predicted. Greater mean age, reduced BMI, hospitalizations in the previous 5 years, and presence of Pseudomonas aeruginosa in sputum cultures were all significantly associated with below-median FEV_1% predicted values. Of the 58 subjects, 21 had been hospitalized (36% of total) and 37 had not. 42 of the subjects were classified as Pseudomonas+ based on sputum microbiology (72.41% of total) and 16 were Pseudomonas-. A greater portion of patients with below-median FEV_1% predicted values had been administered Azithromycin and Tobramycin versus patients with above-median FEV_1% predicted values.

CRP, IL-6, IL-1β, and LBP were significantly correlated with lung function impairment in both univariable and multivariable analysis (Table 2). LPS was significantly related to FEV₁ only in the multivariable analysis. The use of standardized beta-coefficient (i.e. the change in FEV₁% predicted per 1 standard deviation increase in the plasma concentrations of the biomarker) allows for comparison of the beta-coefficients across biomarkers. CRP had the highest standardized beta-coefficient, followed by LBP, IL-6 and IL-1β, suggesting that CRP is most strongly associated with FEV₁. Geometric mean plasma levels of IL-6, IL-1β, CRP, and LBP of patients with below-median FEV_1% predicted values were 3.0 pg/mL, 0.21 pg/mL, 4.6 μg/mL, and 33.2 μg/mL, respectively. These were all significantly higher than in patients with above-median FEV_1% predicted values (1.6 pg/mL, 0.12 pg/mL, 2.0 μg/mL, and 22.9 μg/mL, respectively) (Figure 1). The geometric mean plasma level of GzmB was significantly lower in patients with below-median FEV_1% predicted values (71.6 pg/mL) compared to those with above-median FEV_1% predicted values (166.3 pg/mL).

The geometric mean plasma levels of IL-6, IL-1β, and LPS of hospitalized CF subjects were 3.6 pg/mL, 0.19 pg/mL, and 1.3 ng/mL, respectively. These were all significantly higher than in non-hospitalized subjects (1.7 pg/mL, 0.10 pg/mL, and 0.97 ng/mL, respectively) (Table 3). After adjustment for FEV_1% predicted, BMI and pseudomonal status in sputum, the relationships weakened (Table 3). There was a significant association between plasma IL-6 levels and IL-1β levels (Figure 2). Geometric mean plasma concentrations of CRP, GzmB, CCL18, SP-D, LBP, and sCD14 did not significantly differ between the previously hospitalized and non-hospitalized groups (Table 3). The c-statistic of FEV₁, BMI and pseudomonal status together was 0.797. There was a modest improvement in the c-statistic by adding IL-6, IL-1β, or LPS (0.837, 0.828, and 0.841, respectively).