Relevance of individual bronchial symptoms for asthma diagnosis and control in patients with rhinitis: A MASK‐air study

Abstract Rationale It is unclear how each individual asthma symptom is associated with asthma diagnosis or control. Objectives To assess the performance of individual asthma symptoms in the identification of patients with asthma and their association with asthma control. Methods In this cross‐sectional study, we assessed real‐world data using the MASK‐air® app. We compared the frequency of occurrence of five asthma symptoms (dyspnea, wheezing, chest tightness, fatigue and night symptoms, as assessed by the Control of Allergic Rhinitis and Asthma Test [CARAT] questionnaire) in patients with probable, possible or no current asthma. We calculated the sensitivity, specificity and predictive values of each symptom, and assessed the association between each symptom and asthma control (measured using the e‐DASTHMA score). Results were validated in a sample of patients with a physician‐established diagnosis of asthma. Measurement and Main Results We included 951 patients (2153 CARAT assessments), with 468 having probable asthma, 166 possible asthma and 317 no evidence of asthma. Wheezing displayed the highest specificity (90.5%) and positive predictive value (90.8%). In patients with probable asthma, dyspnea and chest tightness were more strongly associated with asthma control than other symptoms. Dyspnea was the symptom with the highest sensitivity (76.1%) and the one consistently associated with the control of asthma as assessed by e‐DASTHMA. Consistent results were observed when assessing patients with a physician‐made diagnosis of asthma. Conclusions Wheezing and chest tightness were the asthma symptoms with the highest specificity for asthma diagnosis, while dyspnea displayed the highest sensitivity and strongest association with asthma control.


| INTRODUCTION
Asthma guidelines such as the Global INitiative for Asthma (GINA) have improved the knowledge and management of asthma. 1 However, (i) there is a substantial number of asthma misdiagnoses (under/ over-diagnosis) in primary care, partly associated with the under-use of spirometers and under-reporting of symptoms by patients, 2 (ii) many patients diagnosed with asthma are insufficiently monitored and (iii) poor adherence to treatment remains frequent. 3These aspects are all associated with poor control, increased risk of asthma attacks and increased healthcare utilization. 4,5e aspects related to asthma treatment adherence, follow-up, and control may be partly addressed by patient-centered digital solutions allowing the collection of real-world data on asthma symptoms and treatment.The EU Commission has proposed digital solutions, such as wearables and mHealth apps, to engage citizens in the self-management of chronic diseases. 6Thus, the patient perspective, captured by patient-reported outcome measures (PROMs), can strengthen patient-centered care. 7Although various mHealth apps have been developed to support asthma monitoring and management in recent years 8,9 , very few have been developed for both asthma diagnosis and asthma control using validated PROMs.One of the apps, MASK-air ® (Mobile Airways Sentinel networK), is a Good Practice of Directorate-General Health and Food Safety 10 and the only one listed as a Best Practice of Organisation for Economic Cooperation and Development for chronic diseases. 102][13] One such PROM is the Control of Allergic Rhinitis and Asthma Test (CARAT) questionnaire, aiming to assess asthma and AR control together.5][16] CARAT includes three questions on key asthma symptoms (wheezing, dyspnea and chest tightness).mHealth tools and other patient-centered digital solutions can help address the misdiagnosis of asthma not only through the assessment of directly provided daily patient data but also by the use of such data to identify specific variables associated with asthma diagnosis or control.In particular, mHealth-based direct patient data may allow us to assess whether some specific asthma symptoms or symptom patterns are associated with asthma diagnosis or control.This is in line with the recent European Respiratory Society guidelines for the diagnosis of asthma in adults, which, in their conclusion, highlighted that more research should be undertaken on the value of symptoms to predict accurate asthma diagnoses. 17 this study, we used the MASK-air ® PROMs data (CARAT) to assess how individual asthma symptoms (wheezing, dyspnea, chest tightness) were associated with both asthma diagnosis and control.
Knowing which symptom is the most associated with asthma can help reduce the important number of misdiagnoses in the clinical practice, while knowing which symptom has the greatest impact on asthma control is fundamental in the guidance of strategies aiming to improve daily patient management.

| Study design
A full description of the Methods is available in the Online Data Supplement.In this cross-sectional study, we used MASK-air ® data to compare the frequency of occurrence (over the previous month) of five asthma symptoms (dyspnea, wheezing, chest tightness, fatigue and night symptoms) in patients with probable asthma, possible asthma and no evidence of asthma 18

| Settings and participants
MASK-air ® is freely available in 27 countries.In this study, we included data from MASK-air ® users from May 21, 2015 to December 2021, reporting MASK-air ® data in at least three different months.
The users (i) had a self-reported diagnosis of allergic rhinitis and (ii) ranged in age from 16 to 90 years (or lower than 16 years in countries with a lower age of digital consent). 19,20We also included data from participants of the Twinning project who were enrolled during a medical consultation with an asthma specialist. 18Asthma was diagnosed according to GINA, 1 with patients having a pulmonary function test and answering the CARAT questionnaire.Following that consultation, participants were classified as having "current asthma" or "no evidence of (current or past) asthma".

| Ethics
An Independent Review Board (Bohn-Köln) approval was obtained for the MASK-air studies. 21For the Twinning project, additional local review board approvals were obtained, and written consent was provided by the patients.All data were anonymized before the study, and users agreed to the analysis of their data.

| Data sources and variables
The MASK-air ® app comprises a daily monitoring questionnaire assessing (i) the impact of asthma and rhinitis symptoms on a daily basis by means of 0-100 visual analog scales (VASs) (with a higher score corresponding to a higher impact of allergy symptoms) and (ii) the daily use of asthma and rhinitis medication (available from country-specific lists with prescribed and over-the-counter medications). 19The symptom and medication information provided in the MASK-air ® daily monitoring questionnaire allows for the computation of the e-DASTHMA, a 0-100 score assessing the daily control of asthma. 22 addition to the daily monitoring questionnaire, MASK-air ® also includes (although in a non-daily basis) CARAT, a questionnaire assessing the control of allergic rhinitis and asthma in the previous 4 weeks (Table E1). 15

| Data analysis
When responding to the MASK-air ® daily monitoring questionnaire, it is not possible to skip any of the questions, and data are saved to the dataset only after the final answer.This precludes any missing data.All analyses were performed using software R (version 4.0.0).Using a two-step approach, k-means cluster analysis methods were applied to group MASK-air ® users on their probability of having asthma.Obtained clusters subsequently enabled the classification of patients as having "probable asthma", "possible asthma" or "no evidence of asthma" (i.e., rhinitis alone). 18r patients with "probable asthma", "possible asthma" or "no evidence of asthma", using the CARAT questions, 12,15,16 we assessed the frequency of having at least one day per week of (i) dyspnea, (ii) wheezing, (iii) chest tightness, (iv) tiredness/limitations in doing tasks and (v) night symptoms.For each question, we calculated its sensitivity, specificity and predictive values.We also assessed the performance of each question in the discrimination between "possible asthma" and "probable asthma".As each patient could have answered CARAT more than once, we considered, in our main analysis, the first CARAT reported by each patient.We then performed a sensitivity analysis considering (i) all reported CARAT questionnaires or (ii) pre-COVID-19 data only.To validate the obtained results, the sensitivity, specificity and predictive values of each symptom were assessed in a sample of patients in whom the diagnosis of asthma was established by a physician (Twinning participants).
Finally, in patients with probable asthma, we computed the median e-DASTHMA by each category of each CARAT question to assess the most discriminative symptom regarding asthma control.
We considered both the median and maximal e-DASTHMA levels for the 4 weeks before answers to CARAT were provided.

| Demographic characteristics
We included 951 patients with rhinitis (62.8% women; mean � SD age = 38.6 � 13.6), of which 468 (49.2%) had probable asthma, 166 (17.5%) possible asthma and 317 (33.3%) no evidence of asthma.The demographic characteristics of the patients are presented in Tables 1   and 2 and Figure 1).There were 2154 CARAT assessments (Table 3 and Table E2), with 302 indicating controlled CARAT and 1852 indicating uncontrolled complete CARAT.For all these participants, e-DASTHMA information was available.

| Association between individual asthma symptoms and asthma diagnosis
We compared patients with "no evidence of asthma" versus those with "probable or possible asthma" on the presence of each asthma symptom as assessed by CARAT.We examined the first CARAT questionnaire reported by each patient (N = 951, Table 4; Figure 1) and observed that the specificity decreased from wheezing (90.5%) to chest tightness, dyspnea, night symptoms and fatigue (51.1%).On the other hand, the sensitivity was around 50% for wheezing and chest tightness and increased up to 63% for dyspnea.The positive predictive value (PPV) ranged from 71.6% to 90.8% (patients with wheezing had a 90.8% risk of having asthma) and the negative predictive value (NPV) was always lower than 50%.Similar results were observed when all CARAT assessments were reported (Table 4) or when considering pre-COVID-19 data only (Table 4).
In a subsequent analysis, we compared patients with "probable asthma" versus those with "possible asthma" on the presence of each asthma symptom as assessed by CARAT.When considering only the first CARAT questionnaires reported by the patients (Figure 2; Table 5), the specificity decreased from chest tightness (79.5%) and wheezing (77.7%) to dyspnea (64.5%) and fatigue and night symptoms (50.0%).On the other hand, the sensitivity was around 55% for wheezing and chest tightness and increased up to 76.1% for dyspnea.
The PPV ranged from 77.9% to 88.9%.Furthermore, the NPV was always lower than 50%.Similar results were observed when all CARAT assessments were reported (Table 5) or when considering pre-COVID-19 data only (Table 5).

Main analysis (N = 951)
Females-N (%) 597 (62.for night symptoms, 42% for chest tightness and 44% for wheezing (Figure 3).On the other hand, 45% of patients reported an average of more than 2 days per week of dyspnea, 41% reported the same for fatigue, 35% for night symptoms, 28% for chest tightness and 21% for wheezing.

| Validation of results in a sample of patients assessed by a physician
We analyzed 283 Twinning participants, of whom 97 (34.3%) had current asthma and 186 had no evidence of asthma, as assessed by a physician (Table 1).As observed in our main analysis, wheezing was the symptom which displayed the highest specificity (86.0%) and PPV (70.5%), while dyspnea was the symptom associated with the highest sensitivity (76.3%) and NPV (85.6%) (Table 6).The sensitivity and specificity for each symptom were similar to those observed in the main analysis.On the other hand, a lower PPV and a higher NPV were observed in the Twinning study, reflecting differences in the proportion of patients with asthma.

| Association between individual asthma symptoms and asthma control
Dyspnea was the symptom the most strongly associated with e-DASTHMA levels (Table 7).In fact, for dyspnea, the median e-DASTHMA ranged from 5.7 (IQR = 9.5] in patients who had never reported that symptom in the past 4 weeks to 27.8 (IQR = 30.0) in those who reported the symptom every day or almost every day (high effect size = 0.97).Chest tightness and wheezing had a heterogeneous impact on e-DASTHMA (high to no effect).Fatigue and night symptoms had a more consistent effect, with medium effect sizes being observed when comparing median e-DASTHMA levels for patients who had experienced those symptoms every day (or almost) versus those who had never experienced them in the previous 4 weeks.Only dyspnea had a consistent effect on e-DASTHMA levels when all CARAT questionnaires were considered (Table 7).

| DISCUSSION
This is the first study to assess individual asthma symptoms for the diagnosis and control of the disease.Although all symptoms were associated with asthma diagnosis, wheezing was the symptom with the highest specificity and PPV for diagnosis.Dyspnea was the most sensitive symptom associated with the control of asthma as assessed by a novel daily electronic symptom-medication score (e-DASTHMA).

| Strengths and limitations
This study has several limitations.First, the main analysis was not performed in patients with a physician-made diagnosis of asthma.
However, we did not solely rely on self-reported asthma (as asthmaself-reporting tends to be associated with the overestimation or the underestimation of asthma) but rather classified patients as having "probable asthma", "possible asthma" and "no evidence of asthma" based on self-reported asthma status, use of asthma medication and daily asthma symptoms.In a previous study assessing a sample of patients with physician-confirmed diagnosis, we observed that patients clustered as having "probable asthma" had a physician diagnosis of current or past asthma in 92.3% of cases, while patients clustered as having "no evidence of asthma" had a physician diagnosis of "no current asthma" in 90.4% of cases. 18In addition, in this study, we assessed a sample of patients in whom the diagnosis of asthma was established by a physician (Twinning), obtaining consistent results with those of our main analysis.Of note, we did not solely assess patients from asthma clinics with a confirmed diagnosis of asthma since we wanted to have as many patients as possible and in a real-life context.
All assessed patients displayed self-reported rhinitis.Therefore, results are valid only for patients with nasal symptoms, who do however represent a very large proportion of patients with asthma. 23A B L E 4 Comparison between patients classified as having "possible or probable asthma" versus "no evidence of asthma" on the ever occurrence of asthma symptoms as assessed by the Control of Allergic Rhinitis and Asthma Test (CARAT).

Specificity-% (95% CI) PPV-% (95% CI) NPV-% (95% CI)
A. First CARAT assessment by user Shortness of breath/dyspnea a 65.We could not test patients with asthma and without rhinitis as they represent less than 10% of patients with asthma in the MASK-air ® database.
Some of the symptoms examined (e.g.dyspnea, wheezing, chest tightness, and fatigue) are rather nonspecific and may be the expression of different diseases (e.g.asthma and chronic obstructive pulmonary disease (COPD)), and also of the response to treatment in terms of control. 24ditionally, we were not able to assess all asthma symptoms.
For example, we did not have information on the presence of cough and secretions.Nevertheless, secretions appear to be less common than wheezing, dyspnea, chest tightness and sleep disturbances.
Importantly, treatments received prior to or during the time when the patients used the MASK-air app may have influenced the relationship between diagnosis and symptoms.
The tools used have been validated.CARAT is a validated questionnaire of asthma. 16CARAT displays (i) an internal consistency similar to that of the Asthma Control Test (ACT), (ii) a reliability similar to that of the Asthma Control Questionnaire (ACQ) and higher than that of ACT, (iii) a correlation with clinical assessment of asthma control higher than that of ACT (despite being lower than that of ACQ) and (iv) an AUC-ROC for asthma higher than that of the ACQ. 16e-DASTHMA is a daily data-driven asthma control medication score based on MASK-air ® data.It is strongly correlated with daily dyspnea symptoms and moderately correlated with work-and quality-of-life-related comparators.It had high test-retest reliability and displayed moderate-to-high responsiveness.e-DASTHMA was validated in an external cohort of asthma patients enrolled by physicians 25 and was associated with the GINA classification of asthma control. 1 F I G U R E 2 Sensitivity, specificity, positive predictive value and negative predictive value of the occurrence of specific asthma symptoms (Sx; as assessed by the control of allergic rhinitis and asthma test) on asthma diagnosis.
We performed the analysis on the first CARAT questionnaires reported and used all CARAT questionnaires as a sensitivity analysis.

| Interpretation of the results
The first important message is the significance of wheezing in the diagnosis of asthma.Although PPV is influenced by disease prevalence, the high PPV of wheezing for asthma diagnosis is consistent with a study in a secondary care center with a precise asthma diagnosis. 26wever, wheezing lacks sensitivity and 45% of patients with asthma did not report this symptom.Chest tightness displayed a high PPV and a high specificity.It was the symptom showing the highest specificity in differentiating probable from possible asthma.By contrast, dyspnea was the symptom with the highest sensitivity.Despite this result, sensitivity was close to 80%, indicating that approximately one fifth of asthma patients do not experience dyspnea, which likely reflects the good control of the disease in that subset of patients.According to the SpIN and SnOUT rules, 27 a test with high specificity, when positive, helps to rule in the disease.This is the case of wheezing and chest tightness.On the other hand, a test with high sensitivity, when negative, helps to rule out the disease.This may be the case for dyspnea.
The second important message is the significance of dyspnea in asthma control assessment.Dyspnea is a multidimensional, subjective perception of breathing difficulty, commonly seen in patients affected by respiratory diseases, among others. 28Although dyspnea perception may differ between subjects 29 (decreasing with the worsening of asthma, in older adults or in patients with depression 30 ), the present study suggests that it is a cardinal symptom associated with asthma control.In a previous MASK-air ® study, VAS asthma was correlated with VAS dyspnea in patients with different degrees of severity, with a particularly strong correlation being observed in severe asthmatic patients. 31Dyspnea was the symptom which mostly affected the asthma-related quality of life in both mild 26 and severe asthma patients.In a large cross-sectional study encompassing the whole disease severity spectrum, asthma-related quality of life was essentially determined by the level of asthma control. 32These results were also confirmed in a longitudinal study. 33Fatigue and night symptoms were much less specific, and their relationship with asthma control appeared less solid than that of dyspnea and chest tightness.
Future studies may assess whether individual respiratory symptoms can also be useful in the distinction between asthma and other respiratory diseases (namely, COPD).

| Generalizability
This study was conducted with patients from 27 countries.Previous MASK-air ® studies have pointed to a high similarity of outcomes in different countries. 34

| CONCLUSION
Among MASK-air ® users who have rhinitis, wheezing and chest tightness are the symptoms that best predict the presence of asthma (displaying high specificity and PPV), while dyspnea is the symptom that showed the highest sensitivity and the strongest relationship T A B L E 5 Comparison between patients classified as having "probable asthma" versus "possible asthma" on the ever occurrence of asthma symptoms, as assessed by the control of allergic rhinitis and asthma test.with the level of asthma control.These results highlight the importance of thoroughly assessing individual symptoms during both the initial and follow-up assessments of patients with suspected or confirmed asthma.

THE MASK-AIR THINK TANK
The members of the MASK-air think tank are to be acknowledged for having enrolled patients in the study (i.e., for suggesting to their patients to install and use the MASK-air® app).These members include: Marcus Maurer, (1, 2) Rute Almeida, (3, 4) Marek Kulus, (5)   Ignacio J Ansotegui, (6) Fulvio Braido, (7, 8) Victoria Cardona, (9)   Carlos Robalo Cordeiro, (10) Cemal Cingi, (11) Wytske J Fokkens, T A B L E 7 Median e-DASTHMA in patients with probable asthma according to the frequency of specific asthma symptoms as assessed by the control of allergic rhinitis and asthma test.

Never-median (IQR) [N]
Up as assessed by CARAT.Results of this study were validated in a sample of patients in whom asthma diagnosis had been assessed by a physician in the context of a transfer of innovation project (Twinning) of the European Innovation Partnership on Active and Healthy Ageing.

F I G U R E 1
Abbreviation: CARAT, control of allergic rhinitis and asthma test.
Frequency of MASK-air ® users per country.
Abbreviation: SD, Standard-deviation.T A B L E 2enced dyspnea in the previous 4 weeks versus 28% for fatigue, 38%