Small-Airway Function Variables in Spirometry, Fractional Exhaled Nitric Oxide, and Circulating Eosinophils Predicted Airway Hyperresponsiveness in Patients with Mild Asthma.

PURPOSE: Patients with variable symptoms suggestive of asthma but with normal forced expiratory volume in 1 second (FEV1) often fail to be diagnosed without a bronchial provocation test, but the test is expensive, time-consuming, risky, and not readily available in all clinical settings. PATIENTS AND METHODS: A cross-sectional study was performed in 692 patients with FEV1≥80% predicted; normal neutrophils and chest high-resolution computed tomography; and recurrent dyspnea, cough, wheeze, and chest tightness.
RESULTS: Compared with subjects negative for AHR (n=522), subjects positive for AHR (n=170) showed increased FENO values, peripheral eosinophils (EOS), and R5-R20; decreased FEV1, FEV1/Forced vital capacity (FVC), and forced expiratory flow (FEFs) (P≤.001 for all). Small-airway dysfunction was identified in 104 AHR+ patients (61.17%), and 132 AHR- patients (25.29%) (P<0.001). The areas under the curve (AUCs) of variables used singly for an AHR diagnosis were lower than 0.77. Using joint models of FEF50%, FEF75%, or FEF25%-75% with FENO increased the AUCs to 0.845, 0.824, and 0.844, respectively, significantly higher than univariate AUCs (P <0.001 for all). Patients who reported chest tightness (n=75) had lower FEFs than patients who did not (P<0.001 for all). In subjects with chest tightness, the combination of FEF50% or FEF25%-75% with EOS also increased the AUCs substantially, to 0.815 and 0.816, respectively (P <0.001 for all versus the univariate AUCs).
CONCLUSION: FENO combined with FEF50% and FEF25%-75% predict AHR in patients with normal FEV1. FEF25%-75%, FEF50%, or FEF25%-75% together with EOS also can potentially suggest asthma in patients with chest tightness.

Plain Language Summary

Why was the study done?

Patients with variable symptoms suggestive of asthma but with normal forced expiratory volume in 1 second (FEV1, reflect large airway function) often fail to be diagnosed without a bronchial provocation test. But the test is expensive, time-consuming, risky, and not readily available in all clinical settings. Thus, we investigated the predictive values of simple and convenient tests, such as spirometry, fractional exhaled nitric oxide (FENO), and circulating eosinophils for predicting airway hyperresponsiveness (AHR) in patients with mild asthma.

What did the researchers do and find?

We performed a retrospective cross-section study, enrolled 692 patients with normal large airway function (FEV1≥80% predicted) and recurrent dyspnea, cough, wheeze, or chest tightness. We find that small airway dysfunction (FEF25%-75% and FEF50%) do exist in patients with normal FEV1 values but had typical asthmatic symptoms, and FEF25%-75% < 84.4%, or FEF50% < 76.8% combine with FENO > 41 ppb predict positive AHR. FEFs combined with circulating eosinophils also can predict AHR effectively in patients with chest tightness.

What do these results mean?

Combining FEF50% or FEF25%-75% with FENO could forego bronchial provocation test to support asthma diagnosis in patients with normal FEV1 and symptoms suggestive of asthma. FEF25%-75%, FEF50%, or FEF25%-75% together with circulating eosinophils also can potentially suggest asthma in patients with chest tightness.

Introduction

Bronchial asthma (asthma) is a chronic inflammatory airway disease affecting the entire bronchial tree from the large to the small airways (<2 mm diameter).1 The diagnosis is based on recurrent symptoms of dyspnea, cough, wheeze, and chest tightness, as well as reversible airway limitation or airway hyperresponsiveness (AHR).2 Mild asthma is the most common type of asthma, often has normal forced expiratory volume in 1 second (FEV1), representing 50% to 75% of all asthma patients.3 In nearly 90% of the asthmatic patients with normal FEV1 in china (Min Zhang, unpublished paper), the bronchial dilation test is negative, so the bronchial provocation test (BPT) is important for confirming or excluding asthma after detailed medical history and physical examination. However, many hospitals do not perform provocation tests because they are expensive, time-consuming, and entail a risk of severe bronchospasm.2 What’s more, pulmonary function testing should be performed in patients with chronic airway disease during COVID-19 pandemic only if it is needed to guide management and limited to the necessary tests, which usually do not include BPT. Therefore, additional ways are needed to predict AHR safely and to detect patients with mild asthma as early as possible in order to relieve their symptoms and prevent the development of chronic inflammation and airway remodeling.

Small-airway dysfunction exists in mild asthma patient, evaluated by a variety of spirometry and IOS measurements. Lower forced expiratory flow between 25% and 75% (FEF25%-75%) and Forced expiratory flow at 50% of forced vital capacity (FEF50%) are widely used for assessing small-airway function.4–6 Impulse oscillometry (IOS) may also reliably reflect small-airway function and predict clinical asthma outcomes and AHR.7–11 FEV in 3 seconds (FEV3)/FVC is influenced by the airflow velocity in both the central and peripheral airway, normally 95% or greater in adults.12–14

In our previous study of patients with chronic cough, patients with AHR had higher fractional exhaled nitric oxide (FENO), a higher percentage of eosinophils in blood (EOS%), and FEF25%-75% than patients without AHR.4 The combination of FEF25%-75% and FENO increased the area under the curve (AUC) for AHR diagnosis substantially compared with FENO alone. In that study, FEF25%-75% < 78.5% and FENO > 43 ppb strongly predicted positive AHR in Chinese patients with chronic cough.

In the current study, we used FEFs, peripheral airway resistance as the difference between 5 and 20 Hz (R5-R20), reactance at 5 Hz (X5), resonant frequency (Fres), FEV3/FVC, FENO, EOS, and EOS% alone or in combination, as potential predictive variables for the presence of AHR, and extended our cross-sectional study to patients with FEV1 ≥ 80% predicted who had more than 1 typical symptom of asthma, such as variable cough, dyspnea, wheeze, and chest tightness, to confirm the predictive value of small-airway function tests, FENO, and EOS for AHR.

Patients and Methods

Study Design and Subject Selection

A retrospective cross-section of diagnostic data was collected at the initial visit of adult patients with recurrent variable symptoms of dyspnea, cough, wheeze, or chest tightness of at least 8 weeks’ duration who were referred to the Pulmonary Outpatient Clinic of Shanghai General Hospital (China). The patients had to undergo a peripheral blood test, spirometry,15 FENO measurement (NIOX MINO, Aerocrine AB, Solna, Sweden),11 methacholine challenge testing (MCT), and high-resolution computerized tomography (HRCT, GE Medical System; slice thickness 0.625 mm) from September 2016 to January 2020. The MCT was performed with a Jaeger APS Pro system using a Medic-Aid sidestream nebulizer and doubling doses of methacholine (0.015 to 0.48 mg) following the American Thoracic Society/European Respiratory Society recommendations.16

Additional inclusion criteria were age 18–75 years, normal HRCT results, and predicted FEV1 of 80% or greater with spirometric measurement.

Subjects were excluded if they had had respiratory tract infections in the past 8 weeks; peripheral blood test indicating abnormal hemoglobin, platelets, or neutrophils; use of montelukast, long-acting β2-agonists, theophylline, anticholinergic agents, or an inhaled or oral corticosteroid in the previous 4 weeks; or having concomitant severe systemic diseases. Patients who had more than a 10 pack-year smoking history, who currently smoked, or who had quit less than 2 years earlier were also excluded.

Descriptive characteristics, clinical history, results of MCT, spirometry, FENO measurement, IOS (Jaeger Co, Hochberg, Germany),17 EOS, and EOS% were reviewed and analyzed. The FVC, FEV1, peak expiratory flow (PEF), FEF25%, FEF50%, FEF75%, and FEF25%-75% were expressed as percentages of predicted values. FEV1/FVC and FEV3/FVC were expressed as the ratios of the absolute values of the variables. Small-airway dysfunction was identified if 2 of FEF50%, FEF75%, and FEF25%-75% were lower than 80%. IOS variables R5, R20, R5-R20, X5, and Fres were also collected. For bronchial provocation tests, the provocative dose causing a 20% fall in FEV1 (PD20) was recorded, and AHR was defined as present if PD20 was ≤0.48 mg. Associations of PD20 and FENO, EOS, EOS%, FEFs, and IOS were analyzed in patients positive for AHR.

The ethics committee of Shanghai General Hospital, Shanghai Jiao Tong University, approved the protocol. It is a retrospective cross-section study, and the exemption of informed consent will not affect subjects’ rights and welfare. Besides, it is very difficult to contact all the patients of our study. Thus, a waiver of informed consent was given for our study (number 2017KY159). We confirm that the patient data was maintained with confidentiality and complied with the Declaration of Helsinki.

Statistical Analysis

Analyses were performed with SPSS software version 19.0 (SPSS Inc, Chicago, Illinois, USA), except for the ROC contrast estimation, ROC contrast test, and 80/20 split-sample cross-validation, which were performed with SAS Proc LOGISTIC version 9.4 (SAS Institute Inc., Cary, NC, USA). Normality of distribution was checked with the Kolmogorov–Smirnov test. Normally distributed data were expressed as mean ± standard deviation (SD) or 95% Confidence interval (CI). Non-normally distributed data were expressed as median and interquartile ranges (IQR). The coefficient of variance (CV) was calculated. Fisher’s exact test (discontinuous variables) and the Mann–Whitney test (continuous variables) were performed for the intergroup comparisons. The association between different variables was decided by Spearman correlation.

The prediction performance of each variable was measured as the AUC of the ROC. Furthermore, a multiple logistic model of the 2 variables was fitted, and the resultant AUC of this multiple logistic model was used as a measure of the joint prediction performance. Chi-square test proposed by DeLong et al was used to determine whether the multiple logistic model would significantly improve the prediction performance, defined as the AUC, relative to the marginal models.18

MCT was used as the gold standard for defining AHR. The optimal value of the single measurement giving the highest sum of AHR diagnostic sensitivity and specificity was used as a cut-off value.19,20 Positive predictive values (PPV), negative predictive values (NPV), and percentages correctly classified (PCC) were calculated for each cut-off value.21,22 The corresponding odds ratios, CI, and P values were also calculated.

Continuous variables were converted to dichotomous-state variables on the basis of the cut-off values. Subsequently, ROC curves were determined for the joint models with the dichotomous-state variables.

We constructed and examined all models to predict AHR with repeated five-fold cross-validation (5 repeats). The average AUC of 5 different cross-validation models and the whole-model AUC using the entire data set were calculated. The Error Rate equals abs (Average AUC – Whole Model AUC)/(Whole Model AUC). Accurate classification was also calculated for the test subset.

The threshold for statistical significance for all analyses was set at P < 0.05.

Results

Baseline Characteristics

Clinical data from 692 adults were ultimately included. Bronchial provocation tests were positive in 170 patients and negative in 522 patients. Baseline demographics categorized by bronchial provocation test positivity are shown in Table 1. There were no significant differences in age, sex, smoking history, or body mass index (BMI) between the 2 groups. Subjects with AHR had lower percent predicted of FEV1, FEV1/FVC, FEV3/FVC, and PEF (P ≤ 0.001 for all). All FEF values, alone or as ratios of their predicted value to that of FEV1 predicted value, were significantly lower in subjects with AHR than without AHR (P < 0.001 for all).

Table 1

Demographic Data, Spirometric Variables, IOS Variables, and Values for FENO and Peripheral Eosinophils of Patients with Positive or Negative Bronchial Provocation Tests

Variables	CV for All Subjects (%)	Positive Bronchial Provocation Test	CV for BPT+ (%)	Negative Bronchial Provocation Test	CV for BPT− (%)	P value
n	-	170	-	522	-	-
Age, years†	33.79	43.90 (14.56)	33.17	43.80 (14.90)	34.02	0.980
Male, n (%)	-	53 (31.18%)	-	203 (38.89%)	-	0.082
BMI, kg/m2†	14.88	22.94 (2.99)	13.01	23.51 (3.61)	15.37	0.075
Past smoking history (n/%)	-	31 (18.24)	-	97 (18.58)	-	0.204
FVC, % predicted†	11.82	101.10 (10.85)	10.73	101.60 (12.36)	12.17	0.993
FEV1, % predicted†	11.95	99.33 (11.01)	11.08	105.50 (12.49)	11.84	< 0.001
FEV1/FVC, %†	6.72	82.55 (5.51)	6.67	86.87 (5.45)	6.28	< 0.001
FEV3/FVC, %†§	1.98	97.65 (2.26)	2.31	98.47 (1.76)	1.79	0.001
PEF, % predicted†	16.27	90.61 (13.93)	15.38	96.21 (15.65)	16.27	< 0.001
FEF25%, % predicted†	21.63	90.08 (15.87)	17.61	100.80 (22.10)	21.92	< 0.001
FEF50%, % predicted†	25.55	77.26 (17.74)	22.96	98.20 (23.24)	23.66	< 0.001
FEF75%, % predicted†	31.84	74.91 (22.43)	29.94	101.20 (60.72)	29.62	< 0.001
FEF25%-75%, % predicted†	25.51	75.95 (17.64)	23.23	97.01 (22.76)	23.46	< 0.001
Small-airway dysfunction	-	104	-	132	-	< 0.001
R5-R20, kPa·L−1·s†¶	94.65	0.13 (0.11)	88.85	0.10 (0.09)	95.50	< 0.001
X5, kPa·L−1·s†¶	97.54	−0.12 (0.10)	107.93	−0.10 (0.08)	87.30	< 0.001
Fres, L−1·s†¶	42.52	17.53 (9.02)	51.47	14.24 (5.03)	35.29	< 0.001
FENO, ppb‡	94.96	59 (28–98)	75.03	24 (15–45)	86.65	< 0.001
EOS in blood, % ‡	95.46	5.16 (1.70–7.73)	84.47	3.37 (1.20–4.40)	96.58	< 0.001
EOS in blood, cells/μL‡	106.10	370.6 (120–530)	91.90	237.2 (80–302.5)	108.97	< 0.001
PD20, mg‡	130.68	0.079 (0.026- 0.369)	130.68	–	–	–

Notes: Small-airway dysfunction was identified if 2 of the variables FEF50%, FEF75% and FEF25%-75% were lower than 80%. †Mean (SD) values. ‡Median (IQR) values. § n = 333, 92 for positive bronchial provocation, 241 for negative bronchial provocation. ¶ n = 619, 153 for positive bronchial provocation, 466 for negative bronchial provocation. Bold font indicates statistical significance.

Abbreviations: IOS, impulse oscillometry; FENO, fractional exhaled nitric oxide; CV, Coefficient of variance; BPT+, positive bronchial provocation test; BPT−, negative bronchial provocation test; BMI, body mass index; FVC, forced vital capacity; FEV1, forced expiratory volume in 1 second; FEV3, FEV in 3 seconds; PEF, peak expiratory flow; FEF25%, forced expiratory flow at 25% of FVC; FEF50%, FEF at 50% of FVC; FEF75%, FEF at 75% of FVC; FEF25%-75%, FEF at 25% to 75% of FVC; EOS, eosinophils; ppb, parts per billion; PD20, provocative dose causing a 20% fall in FEV in the first second; R5, total airway resistance at 5 Hz; R20, central airway resistance at 20 Hz; R5-R20, peripheral airway resistance as the difference between 5 and 20 Hz; X5, reactance at 5 Hz; Fres, resonant frequency.

Significantly higher levels of R5-R20 and Fres, lower level of X5 were observed in the AHR group (P < 0.001 for all). Furthermore, FENO values, EOS, and EOS% were also dramatically higher in the AHR group (P < 0.001 for all) (Table 1).

Mean values of FEV1, FEV1/FVC, FEF50%, FEF75% and FEF25%-75% were lower in subjects with chest tightness (n = 75) compared with those without this symptom (P = 0.009, 0.008, 0.003, <0.001, and 0.002, respectively). FENO values, EOS, and EOS% values were higher in subjects with than without chest tightness (P = 0.012, 0.001, 0.003, and 0.014, respectively).

Correlation Between PD20 and FENO, EOS, EOS%, FEFs, and R5-R20

FENO, FEF25%, FEF50%, FEF75%, and FEF25%-75% were weakly correlated with PD20 by Spearman analysis (Table 2). No significant correlation between PD20 and FEV3/FVC, EOS, EOS%, R5-R20, X5, or Fres was found. There was no difference in PD20 between subjects with chest tightness (0.065; IQR, 0.220) and those without this symptom (0.089; IQR, 0.349; P = 0.764).

Table 2

Spearman Correlation Between PD20 and Other Variables in BPT+ Subjects

	N	r	P	95% CI
FENO	170	–0.266	0.0005	–0.404 ~ –0.116
EOS	170	–0.054	0.481	–0.208 ~ 0.101
EOS%	170	–0.039	0.609	–0.193 ~ 0.116
FEF25%	170	0.272	0.0003	0.122 ~ 0.409
FEF50%	170	0.213	0.005	0.060 ~ 0.356
FEF75%	170	0.173	0.024	0.018 ~ 0.319
FEF25%-75%	170	0.229	0.003	0.077 ~ 0.371
FEV3/FVC	92	–0.016	0.881	–0.225~ 0.196
R5-R20	153	–0.026	0.754	–0.188 ~ 0.138
X5	153	0.115	0.158	–0.050 ~ 0.273
Fres	153	–0.064	0.433	–0.226 ~ 0.101

Note: Bold font indicates statistical significance.

Abbreviations: PD20, provocative dose causing a 20% fall in FEV in the first second; BPT+, positive bronchial provocation test; CI, confidence interval; FENO, fractional exhaled nitric oxide; EOS, eosinophils; FEF25%, forced expiratory flow at 25% of FVC; FEF50%, FEF at 50% of FVC; FEF75%, FEF at 75% of FVC; FEF25%-75%, FEF at 25% to 75% of FVC; FEV3, FEV in 3 seconds; FVC, forced vital capacity; R5-R20, peripheral airway resistance as the difference between 5 and 20 Hz; X5, reactance at 5 Hz; Fres, resonant frequency.

Diagnostic Accuracy of Single Variables Used for Predicting AHR

We created ROC curves to evaluate the ability of many of the variables to predict positive AHR. In the spirometry measurements, the 2 largest AUCs for a positive AHR diagnosis were 0.763 for FEF25%-75% and 0.762 for FEF50% (Table 3). The FEV1/FVC, EOS%, EOS, and IOS measurements did not give high AUCs for positive AHR diagnosis.

Table 3

Optimal Cut-Off Values and Other Measures of Usefulness for Predicting Airway Hyperresponsiveness

Characteristic Variables	AUC	Cut-Off Values†	Sensitivity %	Specificity %	PPV %	NPV %	PCC %	Odds Ratio	95% CI	P value
FEV1, % predicted	0.649	95.8	47.65	75.86	39.13	81.65	68.93	0.955	(0.939, 0.971)	< 0.001
FEV1/FVC, %	0.713	84.67	66.47	67.62	40.07	86.10	67.34	0.867	(0.837, 0.898)	< 0.001
FEV3/FVC, %	0.616	99.06	67.39	51.87	34.83	80.65	56.16	0.817	(0.726, 0.920)	< 0.001
FEF25%, % predicted	0.656	103.8	84.12	41.38	31.85	88.89	51.88	0.973	(0.963, 0.982)	< 0.001
FEF50%, % predicted	0.762	76.8	58.82	80.46	49.50	85.71	75.14	0.950	(0.939, 0.960)	< 0.001
FEF75%, % predicted	0.745	81.4	66.47	69.92	41.85	86.49	69.08	0.963	(0.955, 0.971)	< 0.001
FEF25%-75%, % predicted	0.763	84.4	70.00	68.97	42.35	87.59	69.22	0.950	(0.940, 0.960)	< 0.001
R5-R20, kPa·L−1·s	0.604	0.88	0.65	100.00	100.00	75.40	75.44	18.961	(3.119, 115.3)	0.001
X5, kPa·L−1·s	0.607	−0.14	43.79	78.76	40.36	81.02	70.11	0.025	(0.003, 0.223)	0.001
Fres, L−1·s	0.634	15.71	56.95	68.32	36.91	82.98	65.53	1.081	(1.046, 1.117)	< 0.001
FENO, ppb	0.748	41	65.29	78.16	49.33	87.37	75.00	1.024	(1.019, 1.030)	< 0.001
EOS in blood, %	0.630	3.4	55.88	66.28	35.06	82.19	63.73	1.130	(1.079,1.182)	< 0.001
EOS in blood, cell/μL	0.638	360	41.76	80.65	41.28	80.96	71.10	4.367	(2.413,7.903)	< 0.001
Platelets, ×10^9/L	0.491	269	80.00	24.71	25.71	79.14	38.30	0.995	(0.995,0.996)	< 0.001

Abbreviations: AUC, area under the curve; PPV, positive predictive values; NPV, negative predictive values; PCC: percentages correctly classified; Odds ratio, odds ratio of characteristic variables for predicting AHR; 95% CI, 95% confidence interval of odds ratio; P value, the p value of the logistic regression test. FEV1, forced expiratory volume in 1 second; FVC, forced vital capacity; FEF25%, forced expiratory flow at 25% of FVC; FEF50%, FEF at 50% of FVC; FEF75%, FEF at 75% of FVC; FEF25%-75%, FEF at 25% to 75% of FVC; R5-R20, peripheral airway resistance as the difference between 5 and 20 Hz; X5, reactance at 5 Hz; Fres, resonant frequency; FENO, fractional exhaled nitric oxide; EOS, eosinophils.

Notes:†The cut-off points were selected by maximizing the sum of sensitivity and specificity. Bold font indicates AUC higher than 0.7.

In patients with chest tightness, the AUCs for positive AHR diagnosis were FEF50% 0.751 (95% CI, 0.637–0.864), FEF75% 0.812 (95% CI, 0.708–0.916), FEF25%-75% 0.763 (95% CI, 0.651–0.875), FENO 0.731 (95% CI, 0.607–0.855), and EOS 0.706 (95% CI, 0.580–0.832).

Diagnostic Accuracy of Small-Airway Function Variables Combined with FENO and Internal Cross-Validation of the Final Models

To determine whether combining measurements would improve AHR prediction, we repeated the ROC analyses for spirometry measurements combined with FENO. The AUC of FEF50% combined with FENO was 0.845 (95% CI, 0.812–0.878), which was significantly higher than the AUC of univariate FEF50% (P < 0.001) (Table 4 and Figure 1). Similarly, the other spirometry measurements also had higher AUCs when combined with FENO than they did alone (Table 4) NPV was ≥85.45% for all of the combinations.

Table 4

Predictive Values of the Combination of Different Variables with FENO in Predicting Airway Hyperresponsiveness

Characteristic Variables	AUC	95% CI of AUC	Sensitivity%	Specificity%	PPV%	NPV%	PCC%	Contrast	95% CI of Contrast	P value
FENO + FEF50%	0.845	(0.812,0.878)	83.53	71.65	48.97	93.03	74.57	0.097†	(0.060,0.135)	< 0.001
FENO + FEF75%	0.824	(0.788,0.859)	72.94	77.01	50.82	89.73	76.01	0.076‡	(0.038,0.114)	< 0.001
FENO + FEF25%-75%	0.844	(0.811,0.876)	80.59	74.14	50.37	92.14	75.72	0.096†	(0.058,0.134)	< 0.001
FENO + FEV1/FVC	0.807	(0.769,0.844)	85.88	61.69	42.20	93.06	67.63	0.059‡	(0.026,0.092)	0.001

Notes: We used the larger of the 2 univariate AUCs to make the comparison; †Contrast, the difference between AUC of FENO and AUC of bivariate model; ‡Contrast, the difference between AUC of FEFs and AUC of bivariate model; P value, contrast’s chi-square test for the significance of the contrast.

Abbreviations: FENO, fractional exhaled nitric oxide; AUC, area under the curve; 95% CI, 95% confidence interval of odds ratio; PPV, positive predictive values; NPV, negative predictive values; PCC, percentages correctly classified; FEF25%, forced expiratory flow at 25% of FVC; FEF50%, FEF at 50% of FVC; FEF75%, FEF at 75% of FVC; FEF25%-75%, FEF at 25% to 75% of FVC; FEV1, forced expiratory volume in 1 second; FVC, forced vital capacity.

Figure 1

ROC curves for the models of FEFs combined with FENO for predicting positive bronchial provocation tests (n = 692). (A) FEF50% combined with FENO. AUCModel = 0.845 (95% CI, 0.812–0.878); AUCFENO = 0.748 (95% CI, 0.702–0.793; P < 0.001, compared with the model); AUCFEF50% = 0.762 (95% CI, 0.721–0.803; P < 0.001, compared with the model). (B) FEF75% combined with FENO. AUCModel = 0.824 (95% CI, 0.788–0.859); AUCFENO =0.748 (95% CI, 0.702–0.793; P < 0.001, compared with the model); AUCFEF75% = 0.745 (95% CI, 0.703–0.786; P < 0.001, compared with the model). (C) FEF25%-75% combined with FENO. AUCModel = 0.844 (95% CI, 0.811–0.876); AUCFENO =0.748 (95% CI, 0.702–0.793; P < 0.001, compared with the model); AUCFEF25%-75% = 0.763 (95% CI, 0.723–0.802; P < 0.001, compared with the model). (D) FEV1/FVC combined with FENO. AUCModel = 0.807 (95% CI, 0.769–0.844); AUCFENO = 0.748 (95% CI, 0.702–0.793; P = 0.001, compared with the model); AUCFEV1/FVC = 0.713 (95% CI, 0.669–0.758).

We then transformed the continuous variables into binary variables according to the cut-off values shown in the tables and reanalyzed the mentioned above ROC curves. The AUCs of FEF50% and FEF25%-75% combined with FENO remained high (Figure 2).

Figure 2

ROC curves of dichotomous state variables of the models of FEFs combined with FENO in predicting positive bronchial provocation tests (n = 692). (A) FEF50% combined with FENO. AUCModel = 0.809 (95% CI, 0.786–0.878); (B) FEF75% combined with FENO. AUCModel = 0.797 (95% CI, 0.768–0.866); (C) FEF25%–75% combined with FENO. AUCModel = 0.802 (95% CI, 0.780–0.874).

In patients with chest tightness, the AUCs of FEF50%, FEF75%, and FEF25%-75% combined with FENO were 0.880 (95% CI, 0.806–0.954), 0.892 (95% CI, 0.812–0.972), and 0.884 (95% CI, 0.805–0.934), respectively (Table 5).

Table 5

Stratified Analysis of Patients with Chest-Tightness: Predictive Values of Combinations of Different Variables with FENO in Predicting Airway Hyperresponsiveness

Characteristic Variables	AUC	95% CI of AUC	Sensitivity%	Specificity%	PPV%	NPV%	PCC%	Contrast	95% CI of Contrast	P value
FENO + FEF50%	0.880	(0.806,0.954)	100.00	60.42	58.70	100.00	74.67	0.129†	(0.022, 0.236)	0.019
FENO + FEF75%	0.892	(0.812,0.972)	70.37	95.83	90.48	85.19	86.67	0.080‡	(–0.001, 0.161)	0.053
FENO + FEF25%-75%	0.884	(0.805,0.964)	88.89	75.00	66.67	92.31	80.00	0.121†	(0.018, 0.224)	0.021
EOS + FEF50%	0.815	(0.716,0.913)	74.07	79.17	66.67	84.44	77.33	0.064‡	(–0.016, 0.144)	0.115
EOS + FEF75%	0.845	(0.750,0.940)	74.07	89.58	80.00	86.00	84.00	0.033	(–0.017, 0.083)	0.197
EOS + FEF25%-75%	0.816	(0.715,0.917)	70.37	85.42	73.08	83.67	80.00	0.053‡	(–0.015, 0.120)	0.125

Notes: We used the larger of the 2 univariate AUCs to make the comparison; †Contrast, the difference between AUC of FENO and AUC of bivariate model; ‡Contrast, the difference between AUC of FEFs and AUC of bivariate model; P value, significance of the contrast by chi-square test. Bold font indicates statistical significance.

Abbreviations: FENO, fractional exhaled nitric oxide; AUC, area under the curve; 95% CI, 95% confidence interval of odds ratio; PPV, positive predictive values; NPV, negative predictive values; PCC, percentages correctly classified; FEF25%, forced expiratory flow at 25% of FVC; FEF50%, FEF at 50% of FVC; FEF75%, FEF at 75% of FVC; FEF25%-75%, FEF at 25% to 75% of FVC; FEV1, forced expiratory volume in 1 second; FVC, forced vital capacity; EOS, eosinophils.

The error rates between the average AUC of 5 different cross-validation models and the whole-model AUC using the entire data set were lower than 0.05 for all chosen variables, indicating that the data model has stable predictive ability for different data sets ( and ).

Diagnostic Accuracy of Small-Airway Function Variables Combined with EOS in Blood

We repeated the ROC analyses for spirometry measurements (FEF50%, FEF25%-75%, FEV1/FVC) combined with EOS (Table 6). The AUCs of the combined models were between 0.734 and 0.786, and all had NPV ≥ 85.08%. The AUCs of FEFs combined with EOS were higher in patients with chest tightness than in the population as a whole (0.815, 0.845, and 0.816, for FEF50%, FEF75%, and FEF25%-75%, respectively; Table 5).

Table 6

Predictive Value of the Combination of Different Variables with EOS in Predicting Airway Hyperresponsiveness

Characteristic Variables	AUC	95% CI of AUC	Sensitivity%	Specificity%	PPV%	NPV%	PCC%	Contrast	95% CI of Contrast	P value
EOS + FEF50%	0.786	(0.748,0.825)	67.06	76.63	48.31	87.72	74.28	0.148	(0.097,0.200)	< 0.001
EOS + FEF25%-75%	0.785	(0.747,0.823)	67.65	75.29	47.13	87.72	73.41	0.147	(0.096,0.199)	< 0.001
EOS + FEV1/FVC	0.734	(0.691,0.777)	56.47	80.84	48.98	85.08	74.86	0.096	(0.045,0.146)	< 0.001

Notes: Contrast, the difference between the AUC of each FEF and the AUC of the bivariate model; we used the larger of the 2 univariate AUCs to make the comparison. P value, significance of the contrast by chi-square test.

Abbreviations: EOS, eosinophils; AUC, area under the curve; 95% CI, 95% confidence interval of odds ratio; PPV, positive predictive values; NPV, negative predictive values; PCC, percentages correctly classified; FEF50%, FEF at 50% of FVC; FEF25%-75%, FEF at 25% to 75% of FVC; FEV1, forced expiratory volume in 1 second; FVC, forced vital capacity.

Discussion

Early diagnosis of asthma is very important, not only to relieve the patient’s symptoms but also to prevent the development of chronic inflammation and airway remodeling. However, early diagnosis is difficult because patients need to satisfy the 2 criteria of variable symptoms and air-flow limitation. This is especially true for mild asthma, which is often undiagnosed because FEV1 is normal and the bronchodilation test has a high false-negative rate. Our data indicating that small-airway dysfunction is present in asthmatic patients with FEV1 ≥ 80% predicted may help to support an early diagnosis in patients with mild asthma. Measurements of small-airway function, including FEF50% and FEF25%-75%, combined with measures of airway inflammation (FENO or EOS) supported the best prediction of positive AHR diagnosis in subjects who had both typical asthma-like symptoms and FEV1 ≥ 80% predicted.

Type II airway inflammation increases both EOS and FENO, which are used in asthma diagnosis and the therapeutic response’s evaluation of anti-asthma drugs.23 Our data also showed that the levels of both EOS and FENO were higher in the AHR-positive group. Previous studies in guinea pigs have shown that NO may itself contribute to AHR, by increasing plasma exudation via its vasodilator effect and by its transformation into peroxynitrite, which induces AHR.24 Furthermore, in our study, a weak correlation between FENO and PD20 in patients with positive AHR was observed (r = −0.266, P = 0.0005), suggesting that FENO is valuable in predicting AHR, which was also noted by Jatakanon et al.25

Currently, FENO is particularly helpful in ruling out asthma. Its cut-off value for predicting asthma ranged from 10.5 to 64 ppb in different studies.26 FENO < 30 ppb has a high specificity (87%) and NPV (93%) for excluding asthma from untreated nonsmoking adults with chronic cough.27 Schleich et al showed that FENO > 34 ppb had a low predictive value (AUC = 0.62) for predicting AHR in patients with suspected asthma.28 Schneider et al illustrated that, in their sample as a whole, asthma could be ruled in at FENO > 71 ppb (PPV, 80%) and ruled out at FENO ≤ 9 ppb (NPV, 82%), with an AUC of 0.656.29 Importantly, when patients with neutrophilic inflammation were omitted, the AUC was 0.745 and asthma could be ruled in at FENO > 31 ppb (PPV, 82%) and ruled out at FENO ≤ 12 ppb (NPV, 81%).29 Our present study showed that FENO > 41 ppb has a sensitivity of 65.29%, specificity of 78.16%, PPV of 49.33%, and NPV of 87.37%. The AUC for predicting AHR was 0.748, regardless of the inflammatory type, which is similar to the AUC from Schneider et al when patients with neutrophilic inflammation were omitted.

In our current study, patients positive for AHR, but with FEV1 in the normal range, had abnormal values of small-airway function variables, obtained with spirometry and IOS. Two-thirds of asthmatic patients with FEV1 ≥ 80% predicted had small-airway dysfunction, and patients with small-airway dysfunction exhibit a greater likelihood of AHR. FEF50%, FEF75%, and FEF25%-75% were weakly correlated with PD20. This might indicate that small-airway dysfunction could be a forerunner of decreased FEV1 and could be used to detect early disease.

We found the 2 most valuable spirometric variables for predicting AHR were FEF25%-75% (AUC = 0.763) and FEF50% (AUC = 0.762) (Table 3). Those two FEFs were strongly correlated and had equivalent value in predicting AHR. However, because both produced AUC < 0.80, using them singly would be insufficient for predicting AHR in patients with suspected asthma. Thus, we combined the FEFs with FENO or EOS to enhance their predictive value for AHR diagnosis. The AUCs of FEFs combined with FENO were significantly higher than those of the univariate AUCs. This suggests that FEFs combined with FENO (2 noninvasive and convenient measurements) can improve the prediction of AHR diagnosis. The cut-off values had certain difference among different studies possibly because we included patients with mild asthma-like symptoms and normal FEV1, who had higher FEFs levels than those with more severe symptoms.5,30–32

One main limitation of FEF is that it depends on FVC and lung capacity.33,34 In contrast to FEV1, FEF25%-75% is not normalized to FVC when assessing air-flow obstruction. Therefore, FEF25%-75% could be artificially low in individuals with restrictive lung or chest bellows disease (eg, obesity) and could therefore overdiagnose asthma. In our study, mean BMI and FVC were in the normal range (23.37 ± 3.477 kg/m2 and 101.5% ± 12%, respectively), and neither variable differed between groups.34,35 Most importantly, all patients in the study had undergone HRCT, therefore guaranteeing that restrictive lung diseases or obesity were excluded and minimizing the possibility of overdiagnosis of asthma in our patients. Furthermore, IOS is a noninvasive and effort-independent alternate test that is done during normal quiet tidal breathing, avoiding the need for any special breathing manoeuvre or any noticeable interference with respiration, and is considered to be more physiological than spirometry36,37 We evaluated the ability of IOS measurement to assess small-airway function and to predict AHR. In our study, R5-R20 alone exhibited poor predictive value for AHR diagnosis, and the AUC was still lower than 0.80 when we combined R5-R20 and FENO. This finding suggests that when combined with FENO, FEFs provide better value than R5-R20 for predicting AHR even though they are less physiologically relevant.

Chest tightness is a symptom of asthma that more likely reflects muscle tightness or physical difficulty with moving air that is sensed through proprioception and not through pain pathways.38–40 In asthmatic patients with normal FEV1 in our study, the most frequent chief complaints were cough and chest tightness rather than wheeze or dyspnea. Relevant clinical subtypes of asthma, “chest tightness variant asthma”41 and “chest pain variant asthma”38,42 have been described in the medical literature. Asthmatic patients who only complain of chest tightness are easily misdiagnosed in clinical practice. We found that decreases of FEV1, FEV1/FVC, FEF25%, FEF50%, FEF75%, and FEF25%-75% were more serious in subjects with than without chest tightness, indicating that small-airway dysfunction may be involved in the mechanism of chest tightness. The joint model of small-airway function variables (FEF50%, FEF75%, or FEF25%-75%) and FENO gave particularly high predictive values for AHR in subjects with chest tightness (all of the AUCs ≥ 0.880). In addition, the joint model of EOS with small-airway function variables (FEF50%, FEF75%, or FEF25%-75%) was highly predictive of AHR in subjects with chest tightness (all of the AUCs ≥ 0.815), which it was not for the population as a whole. Since the cost of peripheral blood cell count is much cheaper than FENO, these tests may provide very economic alternatives for predicting AHR in suspected asthmatics, especially in primary hospitals. The diagnosis of asthma should be strongly considered in patients with lower FEFs, high FENO or high EOS, and the symptom of chest tightness.

The progressive statistical design of this study consisted of several steps. First, the possible influencing variables were found from a Mann–Whitney test. Then, the correlation between the relevant variables and PD20 was determined with Spearman analysis. Through the analysis of AUC, the predictive value of those variables was further verified. The repeatability of our data calculation was shown by an 80/20 split-sample cross-validation. The AUCs in the validation sample were close to those in the whole model.

Some limitations also exist in our current study. Firstly, allergic rhinitis and atopy status were not described in our study, which might influence the diagnostic value of FENO and EOS for BHR. Further comprehensive studies are needed to focus on this issue. Secondly, a larger-scale representative sample of CTVA should be collected to confirm our results. Thirdly, PEF variability measurement needs to be conducted in further prospective study. To overcome the limitations of our current study, larger-scale and multicenter prospective clinical trials should be performed to ensure the integrity of the inspection results.

In conclusion, asthmatic patients suffer from small-airway dysfunction, even though their FEV1 is within the normal range. Patients with small-airway dysfunction exhibit an increased likelihood of having AHR. In order to improve the diagnosis rate of mild asthma and relieve patients’ symptoms as early as possible, we combined 2 simple and noninvasive methods—small-airway function tests and FENO to improve the diagnosis rate of mild asthma. The likelihood of AHR strongly increased with FEF25%-75% <84.4%, FEF50% <76.8%, and FENO >41 ppb. FEV25%-75% and FEF50%, derived from spirometry, could be combined with FENO to support asthma diagnosis in patients with normal FEV1 and symptoms suggestive of asthma, allowing the patient to forego MCT for the diagnosis. FEF25%-75% or FEF50% combined with EOS can also be a very economic method to predict AHR in suspected asthma subjects with chest tightness.