The utility of biomarkers in diagnosis of aspirin exacerbated respiratory disease

Patients with asthma were recruited from the outpatient clinic of the Department of Internal Medicine, Jagiellonian University Medical College, Krakow. All patients were clinically stable without any asthma exacerbations within the 6 weeks preceding the study. All data and sample collection have been previously described [11].

All ATA patients had used aspirin without any adverse effects. Healthy control subjects (HC) without history of asthma, allergy, and any hypersensitivity reactions to NSAIDs were enrolled for comparison of baseline values of inflammatory biomarkers.

Written informed consent was obtained from all participants under a protocol approved by the Jagiellonian University Ethics Review Committee.

Evaluation of asthma was determined by the National Asthma Education and Prevention Program (NAEPP) EPR-3 guidelines [19].

Asthma control was assessed by the Asthma Control Test (ACT) [20].

Spirometry (MasterScreen, Jaeger, Wurzburg, Germany) was carried out at baseline and after administration of 4 puffs of salbutamol. The best of 3 repeatable forced expiratory maneuvers was recorded. The percent predicted values of FEV₁ and FEV₁/FVC were automatically calculated.

Blood eosinophil counts were calculated using a Fuchs-Rosenthal chamber.

Morning urine samples were collected after a 2 h accumulation of urine in the bladder which allows for less variation in urine creatinine levels [10].

Urine creatinine was measured by enzymatic method using automated chemical analyzer COBAS Integra 400 plus (Roche Diagnostics USA, Indianapolis, IN).

Urinary BrTyr was assayed using stable isotope dilution High Performance Liquid Chromatography (HPLC) with on-line electrospray ionization tandem mass spectrometry. First, a solid phase extraction was performed. Briefly 20 μl internal standard composed of 0.5 μM synthetic [¹³C₆]-3-bromotyrosine, 1 mM [¹³C₉, ¹⁵N] tyrosine, and 1 mM creatinine-d₃ (Cambridge Isotope Laboratories) was spiked into 200 μl urine followed by acidification with 1 ml 0.1% formic acid. The acidified urine was loaded to a Biotage® auto solid phase extraction system with 3 ml DSC- 18 column used. DSC-18 column was balanced with 2 × 3 ml methanol and then 2 × 3 ml 0.1% formic acid, washed with 2 × 3 ml 0.1% formic acid, and the product was eluated with 3 ml 0.1% formic acid in 30% methanol. The product was dried under SpeedVacuum, and then resuspended in 100 μl H₂O. Secondly, LC/MS/MS was used to quantify urinary BrTyr. Five microliters of the extraction was analyzed by injection onto a Titan™ C18 UHPLC Column (1.9 μm particle size, L × I.D. 10 cm × 2.1 mm, Supelco) at a flow rate of 0.4 ml min^− 1 using a 2 Shimadzu LC-20 AD Nexera CL pump system, SIL-30 AC MP CL autosampler interfaced with an Shimadzu 8050 mass spectrometer. A discontinuous gradient was generated to resolve the analytes by mixing solvent A (0.2% formic acid in water) with solvent B (0.2% formic acid in methanol) at different ratios starting from 0% B for 3 min, then linearly to 100% B over 3.5 min, then hold for 3 min, and then back to 0% B. [¹³C₉, ¹⁵N₁]-tyrosine was included to simultaneously monitor for potential artificial generation of analyte. 3-bromotyrosine, creatinine, and their respective internal standards [¹³C₆]-3-bromotyrosine and creatinine-d₃ were monitored using electrospray ionization in positive-ion mode with multiple reaction monitoring (MRM) of precursor and characteristic product-ion transitions of m/z 260 → 135, 114 → 44, 266 → 141 and 117 → 47 amu, respectively. The parameters for the ion monitoring were optimized automatically. Nitrogen (99.95% purity) was used as the source, and helium was used as collision gas. Various concentrations of nonisotopically-labeled 3-bromotyrosine were spiked into control urine to prepare the calibration curves for quantification of 3-bromotyrosine. The internal standard [¹³C₆]-3-bromotyrosine was used for quantification as well as to calculate recovery rate of TMAO (which was > 80% based on separate control studies). Under the conditions employed for the assay, no artificial bromination was detected. Results of urinary BrTyr were expressed in nanograms per mg of creatinine.

ULTE₄ was measured in unpurified urine samples by direct enzyme immunoassay (Cayman Chemical, Ann Arbor, MI). Supernatants of freshly collected spot urine samples were stored at − 80°Celsius for not longer than 6 months. After thawing at 4 °C, urine was diluted in phosphate buffered saline (1:10). The measurement was repeated using 1:30 dilution of urine for samples in which the LTE₄ concentration exceeded the uppermost calibrator concentration of the assay (1000 pg/mL). Results of urinary LTE₄ measurements were expressed in picograms per mg of creatinine.

Quantitative characteristics of the study population were described using mean and standard error, while categorical characteristics were described using means and percents. Groups were compared with respect to baseline quantitative characteristics and marker levels among the groups with t-tests and analyses of variance for approximately normally distributed variables, while the Wilcoxon rank-sum test was used with respect to variables demonstrating non-normal distributions. Groups were compared using the likelihood-ratio χ² test with respect to categorical variables. To evaluate uBrTyr, uLTE₄, and blood eosinophils as biomarkers for AERD diagnosis, cut-points were used. A receiver operating characteristic (ROC) analysis was used to identify an optimal cut-point for uBrTyr and uLTE₄. A cut-point for blood eosinophils > 300 cells/μL was based on published data [21, 22]. Logistic regression was used to estimate odds ratios for the identification of AERD (vs ATA) with respect to the dichotomized forms of the biomarkers. The logistic regression analyses were also performed with covariate adjustment for baseline FEV₁%, chronic rhinosinusitis, and any steroid use was performed with the logistic regression modeling. The incremental predictive ability of each marker in the regression model was assessed using areas under the ROC curves estimated with 10-fold cross-validation for the full model and for models excluding markers individually. Correlations among quantitative variables were assessed using Spearman’s rank-sum correlation coefficients, which are suitable for data with any distribution. A p-value less than or equal to 0.05 was defined as statistically significant.

The study included 240 patients with AERD, 226 patients with ATA, and 71 HC. Almost the same population has been previously described [11]. The clinical characteristics of the study subjects are shown in Table 1. Age and gender were distributed significantly differently in asthmatics, i.e., AERD and ATA when compared to (HC), but no difference between AERD and ATA was found. (Table 1). The AERD subjects had worse lung functions and significantly longer asthma duration than the ATA subjects (Table 1). Medication use was different between the AERD and the ATA groups. No difference was found with respect to the ACT scores (Table 1).

Both AERD and ATA groups had significantly higher levels of uBrTyr, uLTE₄, and blood eosinophils than HC (Fig. 1). Furthermore, uLTE₄ levels and blood eosinophils were significantly higher in AERD as compared to ATA (p < 0.0001, p = 0.004, respectively), whereas uBrTyr levels were not significantly different between these two asthmatic phenotypes (p = 0.3406) (Fig. 1).

UBrTyr and uLTE₄ levels correlated with each other both in the AERD [R = 0.160, p = 0.01] and the ATA [R = 0.151, p = 0.02] groups. Blood eosinophils correlated significantly with uLTE₄ levels in both asthma phenotypes [ATA: R = 0.165, p = 0.01; AERD: R = 0.276, p < 0.0001], whereas no such correlation was found with eosinophil counts and uBrTyr levels.

Both uBrTyr and uLTE₄ levels did not correlate with severity of airway obstruction measured by FEV₁% predicted and FEV₁%FVC in either asthma phenotype, while only a borderline correlation was found between FEV₁%FVC and uLTE₄ in ATA [R = 0.140, p = 0.05].

To evaluate if studied biomarkers have the ability to predict an AERD diagnosis, values for each biomarker were dichotomized at cutoffs derived from ROC curve analyses. The cutoff value for uBrTyr was 0.101 ng/mg Cr (specificity 37%, sensitivity 70%), and the cutoff value for uLTE₄ was 800 pg/mg Cr (specificity 81%, sensitivity 50%). A high level of uBrTyr did not confer greater odds of having AERD (OR 1.3; 95% Cl 0.9–2.0, p = 0.15) (Fig. 2). The odds ratio for predicting AERD with respect to blood eosinophils > 300 cells/uL was 1.8 (95% CI 1.2–2.6, p = 0.003) (Fig. 2). A high level of uLTE₄ > 800 pg/mg Cr conferred 4.1-fold odds of having AERD (95% CI 2.7–6.3; p < 0.0001). The combination of high blood eosinophils together with high uLTE₄ levels increased the odds for AERD diagnosis 6.0-fold (95% CI 3.4–10.9, p < 0.0001), with frequency of AERD of 79.6% (74/93) among asthmatics with both characteristics versus 39.5% (77/195) among patients with neither characteristic. Asthmatics with a combination of high blood eosinophil counts, uLTE₄ levels, and uBrTyr levels had 7.1 times greater odds to have AERD than patients with none of the biomarkers elevated (95% CI 3.4–15.8, p < 0.0001) (Fig. 2), with frequency of AERD of 83.1% (59/71) among asthmatics with all three characteristics versus 41.0% (32/78) among patients with none of the characteristics. In the multivariable model with all 3 biomarkers, uBrTyr did not show statistically significant evidence of an additive benefit for predicting AERD when uLTE₄ and blood eosinophils were already taken into account [adjusted odds ratio 1.1, 95% CI 0.8–1.7, p = 0.57].

When performing further covariate adjustment for baseline FEV₁%, chronic rhinosinusitis, and steroid use, the combination of high blood eosinophils together with high uLTE₄ levels increased the odds for AERD diagnosis 3.5-fold (95% CI 1.7–7.4, p = 0.0007), which was smaller than the 6.0-fold estimate without the additional covariate adjustments. Asthmatics with a combination of high blood eosinophil counts, uLTE₄ levels, and uBrTyr levels had covariate-adjusted 7.7 times greater odds to have AERD than patients with none of the biomarkers elevated (95% CI 2.8–24.7, p = 0.0002), which was slightly higher than the 7.1-fold estimate obtained without the additional covariate adjustments. The additive benefit of uBrTyr for predicting AERD when uLTE₄ and blood eosinophils were already taken into account also jumped upward a bit [adjusted odds ratio 1.5, 95% CI 0.9–2.5, p = 0.09] when adjusting for the additional covariates, compared to the 1.1-fold increase estimated without the additional adjustments. The adjustment for steroid use was also performed with respect to only high dose steroid use, but the odds ratio estimates for the biomarkers was virtually identical to those obtained with the adjustment for any steroid use. Using ROC curves, 10-fold cross-validated AUC estimates were 0.82 for the full model and for models excluding either uBrTyr or blood eosinophils, while the exclusion of uLTE₄ yielded an estimate of 0.79. This is not surprising given that among the biomarkers, only uLTE₄ levels yielded a statistically significant association with AERD when adjusting for the other biomarkers and the selected covariates.