Development of a clinical prediction model for cough variant asthma

Abstract

Background:

Cough variant asthma (CVA) is a common cause of chronic cough but remains underdiagnosed due to limited access to bronchial provocation test.

Objective:

To develop a clinical prediction model for CVA based on more accessible indicators.

Design:

A single-center retrospective cohort study.

Methods:

A retrospective cohort of patients with chronic cough from January 2024 to December 2024 was included. The patients were randomly divided into a training set and an internal validation set at a ratio of 7:3. Univariable and multivariable logistic regression analyses were used to identify independent predictors of CVA. A nomogram prediction model was constructed based on these factors. The predictive performance of the nomogram was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA).

Results:

A total of 323 patients with chronic cough were included, with 226 assigned to the training set (45 with CVA, 181 with non-CVA) and 97 to the internal validation set (23 with CVA, 74 with non-CVA). Multivariable logistic regression analysis identified increased eosinophils in induced sputum, elevated peripheral blood eosinophil count (PBEC), raised FeNO₅₀, and reduced maximal mid-expiratory flow (MMEF) as independent predictors of CVA (all p < 0.05). The nomogram model developed based on these indicators demonstrated strong discriminative ability, with area under the ROC curve (AUC) of 0.93 in the training set and 0.92 in the internal validation set. Calibration curves indicated favorable predictive performance of the model. DCA confirmed net clinical benefit.

Conclusion:

The CVA clinical prediction model based on induced sputum cytology, PBEC, FeNO₅₀ and MMEF shows excellent discrimination, accuracy, calibration, and clinical applicability.

Keywords

chronic cough cough variant asthma prediction model

Introduction

Chronic cough is characterized by cough as the sole or predominant symptom lasting more than 8 weeks, with the absence of abnormalities on chest radiographs.¹ Most common etiologies include cough variant asthma (CVA), upper airway cough syndrome (UACS), eosinophilic bronchitis (EB), gastroesophageal reflux-related chronic cough (GERC), and atopic cough (AC).¹ Numerous studies worldwide have indicated that CVA accounts for 32.6% to 52.1% of chronic cough cases, ranking as the most prevalent cause.^2,3 CVA not only significantly impairs patients’ quality of life but also increases healthcare utilization due to delayed or missed diagnoses.⁴ A missed diagnosis of CVA may lead to unnecessary treatments, repeated medical visits, and progression to classic asthma in some cases. Therefore, early and accurate diagnosis of CVA is of paramount importance.

Since CVA lacks typical asthma symptoms such as wheezing and dyspnea, its clinical diagnosis relies on auxiliary examinations including spirometry, bronchial provocation test, and fractional exhaled nitric oxide (FeNO).⁵ However, the clinical application of bronchial provocation test remains limited due to several factors: the risk of inducing bronchospasm, insufficient awareness among physicians, a lack of equipment, medications, and specialized professionals in primary care settings.⁶ Studies have shown that elevated FeNO combined with decreased small airway function can indicate airway hyperresponsiveness in patients with chronic cough, which has a good predictive value for positive bronchial provocation tests.⁷ We previously explored the use of FeNO and small airway parameters in diagnosing CVA and found some utility, their accuracy requires further improvement.⁸ There is great significance in developing an efficient, convenient, and accurate clinical prediction model using noninvasive, widely available and accessible methods. This may promote the precise identification and early intervention of CVA, improve patients’ quality of life and reduce healthcare burden.

This study aims to integrate clinical data from patients with chronic cough in our center, systematically analyze the clinical characteristics and laboratory indicators of CVA patients, and identify key predictors using logistic regression to construct a clinical prediction model. The model’s discriminative ability and calibration will be evaluated through internal validation. Additionally, a visual scoring tool will be developed to assist clinicians in decision-making.

Methods

Subjects and methods

This single-center retrospective study enrolled patients with chronic cough who presented to the Department of Respiratory and Critical Care Medicine of Tongji Hospital from January 2024 to December 2024. All enrolled patients had previously undergone a standardized diagnostic workup for chronic cough, including bronchial provocation test pulmonary function test, FeNO, induced sputum cytology, routine blood test, and serum total immunoglobulin E (tIgE). Additional guideline-based tests were performed for suspected causes: for instance, sinus CT for suspected UACS, multichannel intraluminal impedance-pH monitoring (MII-pH) for suspected GERC, and allergen-specific IgE testing for suspected AC. Clinical data from all these patients were reviewed and collected. They were randomly divided into a training set and an internal validation set at a ratio of 7:3. The training set was used to develop the prediction model, enabling us to identify independent predictors and estimate their relative contributions. The internal validation set was not involved in model development, it was used to assess the model’s performance in an independent but homogeneous cohort, testing its generalizability and uncovering potential overfitting. Based on the final etiological diagnosis, patients were categorized into CVA and non-CVA (NCVA) groups. After comparing clinical characteristics and identifying independent predictors, a prediction model was developed and validated (Figure 1). The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement (Supplemental File 1).⁹

Figure 1.

Flow diagram of the study.

Inclusion criteria

(1) Over 18 years old;

(2) Cough lasting no less than 8 weeks and with a normal chest radiograph;

(3) Forced expiratory volume in 1 second (FEV₁) / forced vital capacity (FVC) >70%, predicted FEV₁ >80%;

Exclusion criteria

(1) Current smokers or ex-smokers of ⩽ 2 years;

(2) Pregnant or lactating patients;

(3) Patients with other respiratory system diseases, such as bronchiectasis, interstitial lung disease, active tuberculosis, lung cancer;

(4) Patients with severe cardiac insufficiency, severe liver and kidney insufficiency, mental and cognitive dysfunction, hearing and communication impairment;

(5) Multiple causes of chronic cough;

(6) Use of angiotensin-converting enzyme inhibitors within the past 3 months;

(7) Data anomalies, including the absence of bronchial provocation test or more than 30% missing data.

Diagnostic criteria

CVA: (1) Chronic cough, often accompanied by a significant night cough; (2) Positive bronchial provocation test; (3) Effective response to antiasthma treatment.

UACS: (1) Chronic cough, predominantly occurring during daytime or after postural changes, and rarely during sleep; (2) Clinical manifestations and history of nasal and/or pharyngeal diseases; (3) Auxiliary examinations supporting the diagnosis of nasal and/or pharyngeal diseases; (4) Cough alleviation after targeted treatment of the underlying condition.

EB: (1) Chronic cough, manifested as irritating dry cough or accompanied by scant sticky sputum; (2) Normal pulmonary ventilation function, no airway hyperresponsiveness, normal PEF variability; (3) Sputum eosinophil proportion ⩾2.5%; (4) Exclusion of other eosinophil-related diseases; (5) Effective response to oral or inhaled corticosteroids.

GERC: (1) Chronic cough, frequently occurring during daytime, and a few patients may have nocturnal cough; (2) Esophageal reflux monitoring showing acid exposure time (AET) > 6% and symptom association probability (SAP) ⩾ 95%; (3) Cough significantly reduced or disappeared after anti-reflux treatment.

AC: Diagnosis is confirmed when criteria (1), (2), (3), (5), and one sub-item of (4) are met:

(1) Chronic cough, mostly irritating dry cough; (2) Normal pulmonary ventilation function and negative bronchial provocation test; (3) No elevation of eosinophils in induced sputum; (4) At least one of the following: a. History of allergic diseases or allergen exposure; b. Positive allergen skin test; c. Elevated total IgE or specific IgE in serum. (5) Effective response to corticosteroids or antihistamines.

Exhaled NO measurement

Exhaled NO was measured using the Nano Coulomb Breath Analyzer (Sunvou-CA2122, Wuxi, China), following the American Thoracic Society/European Respiratory Society (ATS/ERS) recommendations.¹⁰ Subjects were informed about inhaling NO-free air and exhaling via a mouthpiece at two constant flow rates: 50 ml/s, 200 ml/s. FeNO₅₀ and FeNO₂₀₀ were recorded. CaNO was calculated based on the linear model published by ERS: FeNO = CaNO + JawNO/VE. FeNO measurements were performed before spirometric assessments and bronchial provocation tests.

Spirometric measurement

Spirometric assessments were performed with a spirometer (Jaeger, Hoechberg, Germany) in accordance with the specifications and performance criteria recommended in the ATS/ ERS Standardization of Spirometry.¹¹ Small airway dysfunction was defined as at least two of the three parameters falling below 65% of predicted values: maximal mid-expiratory flow (MMEF), forced expiratory flow at 75% of FVC (FEF_75%), and forced expiratory flow at 50% of FVC (FEF_50%).

Histamine bronchial provocation test

Histamine bronchial provocation tests were performed with the Jaeger APS Pro system by using a Medic-Aid sidestream nebulizer, following the recommendations of the ATS/ERS Task force: Standardization of Lung Function Testing.¹² Provocative dose causing a 20% fall in FEV₁ was recorded, and bronchial hyperresponsiveness (BHR) was defined as present if PD₂₀-FEV₁ < 7.8 μmol.

Induced sputum cytology

Induced sputum cytology was performed in accordance with the Chinese national guideline on diagnosis and management of cough.¹ An average of 4.5% saline solution was used for aerosol inhalation.

Statistical analysis

Normally distributed continuous variables were presented as mean ± standard deviation and compared using the t-test. Nonnormally distributed continuous variables were presented as median (Q1-Q3) and compared using the Mann–Whitney U test. Categorical variables were presented as n (%) and analyzed using the chi-square test, while ordinal variables were assessed using the Mann–Whitney U test. Variance inflation factor (VIF) was applied to variables with p < 0.05 in univariable analysis to test for multicollinearity, with VIF < 10 indicating no significant multicollinearity. Variables without multicollinearity in univariable analysis were included in the multivariable logistic regression to identify independent predictors of CVA. Finally, the prediction model was visualized and validated using a nomogram, receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA). The comparison between AUCs was performed using DeLong test. Data analysis was performed using R software 4.4.0 and IBM SPSS Statistics 26.0.

Results

Baseline characteristics

A total of 323 chronic cough patients with single cause were finally included in this study (Table 1). The etiological distribution was as follows: 68 patients with CVA, 90 with GERC, 41 with UACS, 43 with EB, 27 with AC, 37 with refractory chronic cough (RCC), and 17 with other causes. All enrolled patients were randomly divided into a training set (n = 226) and an internal validation set (n = 97) at a ratio of 7:3.

Table 1.

Demographics of patients with chronic cough.

Variables	Training set	Validation set	p Value
N	226	97	–
Etiology, n(%)
CVA	45 (19.91)	23 (23.71)	0.431
GERC	64 (28.32)	26 (26.81)	0.781
UACS	27 (11.94)	14 (14.43)	0.529
EB	31 (13.72)	12 (12.37)	0.744
AC	20 (8.85)	7 (7.22)	0.627
RCC	27 (11.95)	10 (10.31)	0.672
Others	12 (5.31)	5 (5.15)	0.954
Age, years	50 ± 15	49 ± 16	0.862
Female, n(%)	151 (66.81)	52(54.63)	0.001
BMI, kg/m²	24.15 ± 3.73	24.43 ± 3.82	0.535
Duration, months	12 (4–61)	11.50 (3–63)	0.313
Cough symptom score
Daytime	3 (2–4)	3 (2–4)	0.425
Nighttime	1 (1–3)	1 (1–3)	0.367
Atopy, n(%)	86 (38.05)	35 (36.08)	0.62
Eos-sputum, %*	1 (0–3)	1 (0–3)	0.177
Eos-blood, /uL	140 (60–220)	150 (70–320)	0.052
tIgE, IU/mL	34.10 (11.40–96.85)	41.20 (17.10–92.75)	0.162
FeNO₅₀, ppb	16 (11–27)	15 (11–28)	0.985
FeNO₂₀₀, ppb	8 (6–12)	7 (5.00–11.50)	0.394
CaNO, ppb	3.70 (2.40–6.13)	3 (2.15–5.10)	0.089
FEV₁, %predicted	98.38 ± 15.13	98.81 ± 14.17	0.812
FEV₁/FVC, %	80.47 ± 7.48	80.27 ± 6.64	0.819
MMEF, %predicted	80.41 ± 27.35	79.67 ± 24.19	0.817
FEF_75%, %predicted	89.61 ± 21.08	88.65 ± 18.38	0.697
FEF_50%, %predicted	85.46 ± 27.97	85.66 ± 25.09	0.952

Data were presented as mean ± SD or median (Q1–Q3).

Data availability: All variables were available for 226 patients in the training set and 97 patients in the validation set, except for EOS%-sputum (training: n = 201; validation: n = 88).

AC, atopic cough; BMI, body mass index; CaNO, concentration of alveolar nitric oxide; CVA, cough-variant asthma; EB, eosinophilic bronchitis; Eos, eosinophils; FEF_50%, forced expiratory flow at 50% of forced vital capacity; FEF_75%, forced expiratory flow at 75% of forced vital capacity; FeNO, fractional exhaled nitric oxide; FEV₁, forced expiratory volume in 1 second; FVC, forced vital capacity; GERC, gastroesophageal reflux-related chronic cough; MMEF, maximal mid-expiratory flow; ppb, parts per billion; RCC, refractory chronic cough; SD, standard deviation; tIgE, total immunoglobulin E; UACS, upper airway cough syndrome.

In the training set, there were 45 patients in the CVA group and 181 patients in the NCVA group. The characteristics of the two groups are shown in Table 2.

Table 2.

Characteristics of patients in training set.

Variables	CVA	NCVA	p Value
N	45	181	–
Age, years	49 ± 16	51 ± 15	0.545
Female, n(%)	26(57.78)	125(69.06)	0.150
BMI, kg/m²	24.44 ± 3.90	24.07 ± 3.69	0.556
Duration, months	6 (3–48)	13 (4–64)	0.046
Cough symptom score
Daytime	3 (2–4)	3 (2–4)	0.365
Nighttime	1 (1–3)	1 (1–3)	0.576
Atopy, n(%)	32 (71.11)	54 (29.83)	<0.001
Eos-sputum, %*	3 (1–4)	1 (0-2)	<0.001
Eos-blood, /uL	300 (180–570)	100 (50-170)	<0.001
tIgE, IU/mL	98.50 (30–196)	26.70 (9.85–71.95)	<0.001
FeNO₅₀, ppb	40 (31.00–58.50)	13 (10–20)	<0.001
FeNO₂₀₀, ppb	20 (13.50–31.00)	7 (5–9)	<0.001
CaNO, ppb	5.30 (2.85–10.1)	3.50 (2.35–4.85)	0.002
FEV₁, %predicted	94.89 ± 16.85	99.25 ± 14.60	0.084
FEV₁/FVC, %	77.65 ± 8.23	81.17 ± 7.13	0.005
MMEF, %predicted	72.62 ± 26.86	82.35 ± 27.20	0.033
FEF_75%, %predicted	85.45 ± 26.73	90.64 ± 19.38	0.142
FEF_50%, %predicted	77.14 ± 26.81	87.53 ± 27.94	0.025

Data were presented as mean ± SD or median (Q1–Q3).

Data availability: All variables were available for all 226 patients, except for EOS%-sputum (CVA: n = 39; NCVA: n = 162).

BMI, body mass index; CaNO, concentration of alveolar nitric oxide; CVA, cough-variant asthma; Eos, eosinophils; FEF_50%, forced expiratory flow at 50% of forced vital capacity; FEF_75%, forced expiratory flow at 75% of forced vital capacity; FeNO, fractional exhaled nitric oxide; FEV₁, forced expiratory volume in 1 second; FVC, forced vital capacity; MMEF, maximal mid-expiratory flow; NCVA, non-cough-variant asthma; ppb, parts per billion; SD, standard deviation; tIgE, total immunoglobulin E.

Identification of independent predictors for CVA

Univariable logistic regression analysis was performed on variables with statistically significant differences between the two groups in the training set. The incidence of atopy, percentage of eosinophil in induced sputum, PBEC, total IgE (tIgE), FeNO₅₀, FeNO₂₀₀, CaNO, FEV₁/FVC%, MMEF, and FEF_50% were all statistically significant (Table 3). Collinearity testing revealed that VIF of MMEF, FEF_50%, FeNO₅₀, FeNO₂₀₀, and CaNO was > 10, so FEF_50%, FeNO₂₀₀, and CaNO were excluded from subsequent multivariable analysis. The remaining seven variables were included in the multivariable logistic regression analysis. The results identified percentage of eosinophil in induced sputum, PBEC, FeNO₅₀, and MMEF as independent predictors for CVA (p < 0.05).

Table 3.

Logistic regression analysis.

Variables	Univariate		Multivariate
Variables	p Value	OR (95%CI)	p Value	OR (95%CI)
Duration, months	0.572	0.999 (0.994–1.003)
Atopy	<0.001	5.789 (2.821–11.881)	0.175	2.503 (0.665–9.417)
Eos-sputum, %	<0.001	1.429 (1.206–1.693)	0.003	1.566 (1.171–2.095)
Eos-blood, /ul	<0.001	1.006 (1.004–1.008)	0.008	1.004 (1.001–1.007)
tIgE, IU/mL	0.002	1.002 (1.001–1.003)	0.740	1.000 (0.997–1.002)
FeNO₅₀, ppb	<0.001	1.149 (1.102–1.198)	<0.001	1.163 (1.103–1.226)
FeNO₂₀₀, ppb	<0.001	1.330 (1.211–1.460)
CaNO, ppb	<0.001	1.139 (1.066–1.216)
FEV₁/FVC, %	0.006	0.941 (0.900–0.983)	0.569	0.955 (0.814–1.120)
MMEF, %predicted	0.034	0.986 (0.974–0.999)	0.006	0.962 (0.936–0.989)
FEF_50%, %predicted	0.027	0.986 (0.973–0.998)

CaNO, concentration of alveolar nitric oxide; Eos, eosinophils; FEF_50%, forced expiratory flow at 50% of forced vital capacity; FeNO, fractional exhaled nitric oxide; FEV₁, forced expiratory volume in 1 second; FVC, forced vital capacity; MMEF, maximal mid-expiratory flow; tIgE, total immunoglobulin E.

Diagnostic value of individual predictors for CVA

ROC analysis was performed on the above independent predictors of CVA (Table 4). FeNO₅₀ and PBEC had good diagnostic value for CVA, with areas under the curve (AUC) of 0.89 and 0.84, respectively, and there was no difference between these two AUCs (DeLong test, p = 0.14). FeNO₅₀ showed a high sensitivity of 91.11%. In contrast, percentage of eosinophil in induced sputum had the lowest diagnostic value, with an AUC of 0.69.

Table 4.

Optimal cutoff values for the prediction of CVA.

Parameter	Cutoff	AUC (95%CI)	Sensitivity	Specificity	PPV	NPV	PLR	NLR
Eos-blood	150	0.84(0.79–0.89)	82.22	71.82	42.00	94.20	2.92	0.25
Eos-sputum	2	0.69(0.62–0.75)	51.11	79.01	37.70	86.70	2.43	0.62
FeNO₅₀	26	0.89(0.85–0.93)	91.11	81.77	55.40	97.40	5.00	0.11
MMEF	69.70	0.76(0.70–0.82)	64.44	81.22	46.00	90.20	3.43	0.44

Sensitivity, specificity, PPV and NPV are in %; the cutoff point of MMEF is in %predicted; the cutoff point of FeNO₅₀ is in ppb; the cutoff point of Eos-blood is in cells/uL; and cutoff point of Eos-sputum is in %.

AUC, area under the curve; Eos, eosinophils; FeNO, fractional exhaled nitric oxide; MMEF, maximal mid-expiratory flow; NLR, negative likelihood ratio; NPV, negative predictive value; PLR, positive likelihood ratio; PPV, positive predictive value.

Construction and evaluation of clinical prediction model

Based on the results of multivariable logistic regression analysis, the percentage of eosinophils in induced sputum, PBEC, FeNO₅₀, and MMEF were included to construct a nomogram prediction model for CVA (Figure 2). The probability of CVA in chronic cough patients can be estimated by calculating the total score from the vertical line of the variable to the point axis. Regression equation: logit(Risk) = −4.602386581+0.392852951*EOS_sputum+0.003186175*EOS_blood−0.026163156*MMEF+0.136370733*FeNO₅₀

Figure 2.

Nomogram of CVA prediction model.

The model showed excellent predictive accuracy for CVA, with the AUC of 0.933(95%CI, 0.901–0.965), the sensitivity of 0.978, and the specificity of 0.818 (Figure 3(a)). The calibration curve shows that the apparent line and the bias-corrected line are basically coincident, with a mean absolute error of 0.022, suggesting excellent consistency between the model predictions and the actual observations (Figure 4). Decision curve analysis revealed that the prediction model provided a net clinical benefit when the threshold probability was between 1% and 95%, using the prediction model to guide the diagnosis of CVA yields a higher net benefit compared to strategies of “diagnose all” or “diagnose none,” demonstrating the model’s applicability across a wide spectrum of clinical decision-making scenarios (Figure 5). Furthermore, the model was validated using the internal validation set, which yielded an AUC of 0.923 (95%CI, 0.869–0.976), confirming its robust predictive performance (Figure 3(b)).

Figure 3.

ROC curve of CVA prediction in the (a) training set and (b) validation set.

Figure 4.

Calibration curve of CVA prediction in the training set.

Figure 5.

Decision curve analysis for the CVA prediction in the training set.

Simplified clinical prediction model

Given the potential risks and limited availability of induced sputum cytology, we further constructed a simplified prediction model based on PBEC, FeNO₅₀, and MMEF (Figure 6). The model showed good predictive value for CVA, with the AUC of 0.878(95%CI, 0.812–0.944), the sensitivity of 0.677, and the specificity of 0.908 (Figure 7(a)). The simplified model was validated in the internal validation set (Figure 7(b)), with the AUC of 0.871(95%CI, 0.791–0.952). Figure 8 showed the calibration of the model. The predicted probabilities of the model are generally consistent with the observed risks, indicating good model performance. Figure 9 presented the decision curve analysis. Across a wide range of clinically reasonable threshold probabilities, the net benefit of using this model to guide diagnosis is higher than the strategies of “diagnosing all” or “diagnosing none,” suggesting that the model has good clinical utility.

Figure 6.

Nomogram of simplified CVA prediction model.

Figure 7.

ROC curve of simplified CVA prediction in the (a) training set and (b) validation set.

Figure 8.

Calibration curve of simplified CVA prediction in the training set.

Figure 9.

Decision curve analysis for the simplified CVA prediction in the training set.

Regression equation: logit(Risk) = −4.314400502+0.003486655*EOS_blood−0.017734247*MMEF+0.131192936*FeNO₅₀

Discussion

This study integrated clinical characteristics and laboratory parameters of patients with chronic cough, identifying percentage of eosinophil in induced sputum, PBEC, FeNO₅₀, and MMEF as independent predictors of CVA. A nomogram prediction model based on these indicators demonstrates favorable clinical utility.

Our previous study found that FeNO and small airway parameters have certain value in the diagnosis of CVA, showing high specificity but low sensitivity, consistent with findings reported by Chen et al.^8,13,14 Although FeNO and small airway parameters exhibit good predictive value for airway hyperresponsiveness,⁷ their predictive accuracy for CVA needs to be further improved. Through systematic analysis of clinical characteristics in CVA patients, four independent predictors which were identified in this study closely align with the core pathophysiological features of CVA: type 2 airway inflammation and small airway dysfunction. At the same time, four predictors included in the model are themselves correlated with airway hyperresponsiveness, there may be a minimal incorporation bias.

At present, commonly used type 2 inflammation biomarkers in the clinical assessment of respiratory diseases include PBEC, percentage of eosinophil in induced sputum, FeNO, tIgE, and allergen-specific IgE (sIgE). PBEC is simple and feasible, and correlated with the eosinophil count in induced sputum. It also has moderate predictive value in identifying increased eosinophils in sputum,¹⁵ and can guide the diagnosis and treatment of asthma, chronic obstructive pulmonary disease, and chronic cough, especially in community clinics and secondary hospitals.^16,17 However, the specific cut-off value of PBEC remains inconsistent. A count ⩾150/μL may indicate an eosinophilic phenotype or type 2 inflammation endotype and can serve as a reference for evaluating anti-inflammatory treatment response and the efficacy of biologic therapies. Some studies have shown that in asthma patients, an eosinophil count > 300/μL reflects eosinophilic airway inflammation with sensitivity over 70% and specificity over 90%.¹⁸ In this study, the optimal cutoff value of blood eosinophil count for diagnosing CVA is 150/μL, with an AUC of 0.84, which is helpful for clinicians to discover potential CVA clues based on this most common test. The proportion of eosinophils in induced sputum is one of the most direct indicators for assessing airway inflammation in asthma. It is used for asthma classification, prediction of the response to glucocorticoid treatment, and evaluating acute exacerbation risk. It also serves as a noninvasive method for diagnosing chronic cough etiology and assessing airway inflammation, with good safety and tolerability. Elevated sputum eosinophils are essential for diagnosing EB and can also assist in diagnosing CVA. In China, sputum eosinophil percentage ⩾2.5% is considered the standard for increased sputum eosinophils. Most asthma patients have an increased proportion of eosinophils in induced sputum,¹⁹ which is related to asthma symptoms, and anti-inflammatory treatment can reduce the sputum eosinophil count. However, the sputum EOS% of CVA patients is lower than that of typical asthma patients.²⁰ In this study, the cut-off value of sputum eosinophil percentage for diagnosing CVA was 2%, with a relatively low AUC. This may be influenced by the inclusion of EB patients in the NCVA group. Additionally, the technical demands of induced sputum cytology, time requirements, and difficulties in obtaining adequate sputum samples from patients with dry cough limit its widespread application. FeNO is an internationally recognized biomarker of type 2 airway inflammation. It not only reflects the level of airway inflammation but also predicts therapeutic response to inhaled corticosteroids and type 2 inflammation-related monoclonal antibodies. Its noninvasive and convenient nature makes it a commonly used test in respiratory diseases. CVA patients exhibit higher FeNO than those with chronic cough due to other causes, enabling FeNO to identify CVA among chronic cough patients.⁸ A study on adult chronic cough patients in China showed that FeNO ⩾31.5 ppb indicates a high possibility of corticosteroid-responsive cough, and FeNO ⩾33.5 ppb suggests a high possibility of CVA.²¹ However, the sensitivity is relatively low, only 54.0% and 69.6%, respectively, potentially leading to missed diagnoses if these values are used as cutoffs. In previous studies that used FeNO alone for the diagnosis of CVA, although the AUC could reach above 0.8, the sensitivity and specificity were not simultaneously higher than 80%.^8,22 In contrast, the nomogram model developed in this study, which incorporates multiple indicators including FeNO, significantly improves predictive accuracy.

Small airway dysfunction is prevalent across all severity levels of bronchial asthma. A global multicenter study revealed that small airway dysfunction exists in 91% of asthma patients.²³ Patients with small airway dysfunction are more likely to have airway hyperresponsiveness, and FEF_50%, FEF_75%, and MMEF are correlated with PD₂₀.²⁴ Although CVA patients typically have normal pulmonary ventilation function, impaired small airway function is frequently observed. It was reported that when small airway dysfunction was defined as at least two of the parameters MMEF, FEF_50%, or FEF_75% falling below 65% of predicted values, more than 60% of CVA patients have small airway dysfunction.²⁵ Our previous study also demonstrated that the incidence of small airway dysfunction in CVA patients is significantly higher than that in other chronic cough patients. The AUC of small airway parameters FEF_50%, FEF_75%, and MMEF for CVA is about 0.75, and combining FeNO with these parameters enhanced predictive performance.⁸ In the current study, MMEF achieved an AUC of 0.76 for CVA diagnosis, consistent with previous findings, further validating the importance of small airway parameters and supporting the necessity of including MMEF in the prediction model.

Although serum total IgE is a recognized type 2 inflammation biomarker, it was not identified as an independent predictor in this study. IgE plays a key role as a mediator in type 2 immune responses, particularly in initiating allergic reactions and inflammatory processes mediated by mast cells or basophils. Serum tIgE are influenced by various factors such as age, gender, season, and allergen exposure. High tIgE could be observed in various atopic diseases, but nearly half of atopic patients exhibit normal tIgE levels.²⁶ Previous studies have shown that the serum tIgE of CVA patients is significantly lower than that of classic asthma patients, and 55.6% of CVA patients have no clear allergens.²⁷ Among chronic cough patients, elevated tIgE may occur not only in some CVA cases but also in AC patients and those with other etiologies accompanied by allergic rhinitis or atopic dermatitis. In addition, FeNO₂₀₀ and CaNO were not included in the final model. FeNO₅₀ reflects the inflammation of the trachea and bronchi, while FeNO₂₀₀ indicates the eosinophilic inflammation in the peripheral small airways, and CaNO is calculated based on a specific model formula, reflecting the degree of airway inflammation in the alveolar region. Due to the strong collinearity among them, and considering the higher accessibility of FeNO₅₀ as well as the inclusion of small airway parameters in the model, FeNO₅₀ was retained in the model construction.

In summary, this clinical prediction model is not merely a combination of related indicators, but establishes a multidimensional framework for CVA prediction by integrating three distinct dimensions: inflammatory cells (sputum and blood eosinophils), inflammatory mediators (FeNO), and functional impairment (small airway function). The model achieved an AUC of 0.933, demonstrating superior predictive performance compared with individual parameters. Importantly, it provides clinicians with a practical, low-risk, and efficient tool for CVA diagnosis, potentially reducing the need for bronchial provocation tests, which are more invasive, time-consuming, and associated with higher risk in patients with airway hyperresponsiveness. By relying primarily on accessible and minimally invasive measurements, the model has greater applicability in routine clinical practice and in primary care settings. At the same time, given the potential risks and limited availability of induced sputum cytology, we further developed a simplified model using readily available and minimally invasive variables: blood eosinophil count, FeNO, and MMEF. This simplified model retained good predictive performance, with an AUC of 0.878, slightly lower than the full model, indicating it can serve as a practical and accessible tool for CVA screening, particularly in primary care or resource-limited settings. While the full model provides a more comprehensive assessment by including sputum eosinophils, the simplified model demonstrates that accurate prediction is achievable using indicators that are safer, easier to obtain, and more widely implementable. Together, these findings suggest a flexible approach: clinicians can select the model that best fits their available resources and patient population.

There are some limitations in this study. As a single-center retrospective study, it is subject to inherent biases, and the relatively limited sample size may affect the generalization ability of the model. Although the indicators used in this model are more accessible than bronchial provocation test, induced sputum cytology examination remains challenging in some hospitals, and the availability of FeNO testing equipment needs further enhanced. Although there was a statistically difference in the gender distribution between the training set and the validation set in this study, gender was not selected as an independent predictor in the prediction model. This imbalance had a relatively small impact on the model’s discriminatory ability. Future multicenter studies should further verify the gender universality of the model. External validation through larger and multicenter prospective studies is warranted in the future.

Conclusion

This study successfully developed and preliminarily validated a nomogram prediction model for CVA based on the percentage of eosinophils in induced sputum, PBEC, FeNO₅₀ and MMEF. The model demonstrates favorable predictive performance and calibration, indicating strong potential for clinical application. By closely aligning with the type 2 inflammation and small airway dysfunction characteristics of CVA, and integrating multidimensional noninvasive indicators, this model provides clinicians with an intuitive and convenient diagnostic tool. It contributes to improved diagnostic efficiency for CVA, particularly in resource-limited settings or where bronchial challenge tests are not feasible.

Supplemental Material

sj-docx-1-tar-10.1177_17534666261450064 – Supplemental material for Development of a clinical prediction model for cough variant asthma

Supplemental material, sj-docx-1-tar-10.1177_17534666261450064 for Development of a clinical prediction model for cough variant asthma by Haodong Bai, Baiyi Yi, Tongyangzi Zhang, Yiqing Zhu, Wanzhen Li, Shengyuan Wang, Xianghuai Xu and Li Yu in Therapeutic Advances in Respiratory Disease

Footnotes

Acknowledgements

None.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Tongji Hospital Affiliated to Tongji University (K-2025-003) on January 7th, 2025, with the need for written informed consent waived.

Consent for publication

Not applicable.

Author contributions

Haodong Bai: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Visualization; Writing – original draft.

Baiyi Yi: Data curation; Methodology; Software; Writing – original draft.

Tongyangzi Zhang: Conceptualization; Methodology; Resources; Software

Yiqing Zhu: Investigation; Resources.

Wanzhen Li: Data curation; Investigation.

Shengyuan Wang: Methodology; Resources; Software.

Xianghuai Xu: Conceptualization; Funding acquisition; Methodology; Project administration; Resources; Supervision; Validation; Writing – review & editing.

Li Yu: Conceptualization; Funding acquisition; Methodology; Project administration; Resources; Supervision; Validation; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the National Natural Science Foundation of China (No. 82270114 and 82570146), the Seed Program for Research and Transformation of Medical New Technologies of the Shanghai Municipal Health Commission (2025ZZ1028), the Key Supported Discipline of Health System in Shanghai (2023ZDFC0302), Shanghai Leading Talent Program of Eastern Talent Plan (leading talent program) (BJWS2025069), the Program of Shanghai Municipal Health Commission Clinical Research (No. 20234Y0190, 20244Y0147 and 20254Y0012), Medical Research Project of China Medical and Health Development Foundation (BJ2024JCHX002).

Competing interests

The authors declare that there is no conflict of interest.

Availability of data and materials

The datasets used in the current study are available from the corresponding author* on reasonable request.

ORCID iDs

Wanzhen Li

Shengyuan Wang

Xianghuai Xu

Li Yu

Supplemental material

Supplemental material for this article is available online.

References

Asthma Group of Chinese Thoracic Society. Chinese national guideline on diagnosis and management of cough (2021). Chin J Tuberc Respir Dis 2022; 45: 13–46.

Lai

Chen

Lin

, et al. A prospective, multicenter survey on causes of chronic cough in China. Chest 2013; 143: 613–620.

Ding

Wen

, et al. Changing etiological frequency of chronic cough in a tertiary hospital in Shanghai, China. J Thorac Dis 2019; 11: 3482–3489.

Lai

Chen

, et al. Prevalence and burden of chronic cough in China: results from a population-based survey. BMC Pulm Med 2025; 25: 234.

Chinese Thoracic Society. Guidelines for the prevention and management of bronchial asthma (2024 edition). Chin J Tuberc Respir Dis 2025; 48: 208–248.

Asthma Group of Chinese Thoracic Society. Chinese expert consensus on the clinical applications of bronchial provocation test (2024 edition). Natl Med J China, 2024; 104: 2031–2040.

Bao

Zhang

, et al. The value of fractional exhaled nitric oxide and forced mid-expiratory flow as predictive markers of bronchial hyperresponsiveness in adults with chronic cough. J Allergy Clin Immunol Pract 2018; 6: 1313–1320.

Bai

Shi

, et al. A comparative study on the value of lower airway exhaled nitric oxide combined with small airway parameters for diagnosing cough-variant asthma. Ther Adv Respir Dis 2023; 17: 17534666231181259.

von Elm

Altman

Egger

, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet 2007; 370: 1453–1457.

10.

ATS/ERS recommendations for standardized procedures for the online and offline measurement of exhaled lower respiratory nitric oxide and nasal nitric oxide, 2005. Am J Respir Crit Care Med 2005; 171: 912–930.

11.

Graham

Steenbruggen

Miller

, et al. Standardization of Spirometry 2019 Update. An Official American Thoracic Society and European Respiratory Society Technical Statement. Am J Respir Crit Care Med 2019; 200: e70-e88.

12.

Miller

Enright

PL.

PFT interpretive strategies: American Thoracic Society/ European Respiratory Society 2005 guideline gaps. Respir Care 2012; 57: 127–133; discussion 133-135.

13.

Chen

Zeng

, et al. Diagnostic value of FeNO and MMEF for predicting cough variant asthma in chronic cough patients with or without allergic rhinitis. J Asthma 2021; 58: 326-333.

14.

Feng-Jia

Xin-Yan

Geng-Peng

, et al. Validity of fractional exhaled nitric oxide and small airway function indices in diagnosis of cough-variant asthma. J Asthma 2018; 55: 750–755.

15.

Fang

Liang

, et al. Diagnostic accuracy of blood eosinophils in comparison to other common biomarkers for identifying sputum eosinophilia in patients with chronic cough. World Allergy Organ J 2023; 16: 100819.

16.

Caspard

Ambrose

Tran

, et al. Associations between individual characteristics and blood eosinophil counts in adults with asthma or COPD. J Allergy Clin Immunol Pract 2020; 8: 1606–1613.e1601.

17.

Korevaar

Westerhof

Wang

, et al. Diagnostic accuracy of minimally invasive markers for detection of airway eosinophilia in asthma: a systematic review and meta-analysis. Lancet Respir Med 2015; 3: 290–300.

18.

Wagener

de Nijs

Lutter

, et al. External validation of blood eosinophils, FE(NO) and serum periostin as surrogates for sputum eosinophils in asthma. Thorax 2015; 70: 115–120.

19.

Petsky

Cates

Lasserson

, et al. A systematic review and meta-analysis: tailoring asthma treatment on eosinophilic markers (exhaled nitric oxide or sputum eosinophils). Thorax 2012; 67: 199–208.

20.

Gao

, et al. Inflammatory subtypes in classic asthma and cough variant asthma. J Inflamm Res 2020; 13: 1167–1173.

21.

Chen

Luo

, et al. Validity of fractional exhaled nitric oxide in diagnosis of corticosteroid-responsive cough. Chest 2016; 149: 1042–1051.

22.

Liu

Wang

Yao

, et al. Value of exhaled nitric oxide and FEF(25-75) in identifying factors associated with chronic cough in allergic rhinitis. Allergy Asthma Immunol Res 2019; 11: 830–845.

23.

Postma

Brightling

Baldi

, et al. Exploring the relevance and extent of small airways dysfunction in asthma (ATLANTIS): baseline data from a prospective cohort study. Lancet Respir Med 2019; 7: 402–416.

24.

Bao

Zhang

Yin

, et al. Small-airway function variables in spirometry, fractional exhaled nitric oxide, and circulating eosinophils predicted airway hyperresponsiveness in patients with mild asthma. J Asthma Allergy 2021; 14: 415–426.

25.

Jiang

, et al. Small airway dysfunction in cough variant asthma: prevalence, clinical, and pathophysiological features. Front Physiol 2021; 12: 761622.

26.

Zhao

Zhu

, et al. Age, sex, and symptom-dependent variations in total IgE and eosinophils in atopic patients: a five-year retrospective study. J Asthma Allergy 2025; 18: 915–926.

27.

Lai

Zhan

, et al. Clinical and inflammatory characteristics of the chinese APAC cough variant asthma cohort. Front Med (Lausanne) 2021; 8: 807385.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.03 MB