Accepted Manuscript /j.jaci Reference: YMAI To appear in: Journal of Allergy and Clinical Immunology

Transcription

1 Accepted Manuscript Diagnostic accuracy of fractional exhaled nitric oxide in predicting cough variant asthma and eosinophilic bronchitis in adult patients with chronic cough: A systematic review and meta-analysis Woo-Jung Song, MD, Hyun Jung Kim, MPH, PhD, Ji-Su Shim, MD, Ha-Kyeong Won, MD, Sung-Yoon Kang, MD, Kyoung-Hee Sohn, MD, Byung-Keun Kim, MD, Eun- Jung Jo, MD, Min-Hye Kim, MD, PhD, Sang-Heon Kim, MD, PhD, Heung-Woo Park, MD, PhD, Sun-Sin Kim, MD, PhD, Yoon-Seok Chang, MD, PhD, Alyn H. Morice, MD, FRCP, Byung-Jae Lee, MD, PhD, Sang-Heon Cho, MD, PhD PII: DOI: S (17) Reference: YMAI /j.jaci To appear in: Journal of Allergy and Clinical Immunology Received Date: 3 June 2016 Revised Date: 28 October 2016 Accepted Date: 2 November 2016 Please cite this article as: Song W-J, Kim HJ, Shim J-S, Won H-K, Kang S-Y, Sohn K-H, Kim B-K, Jo E-J, Kim M-H, Kim S-H, Park H-W, Kim S-S, Chang Y-S, Morice AH, Lee B-J, Cho S-H, Diagnostic accuracy of fractional exhaled nitric oxide in predicting cough variant asthma and eosinophilic bronchitis in adult patients with chronic cough: A systematic review and meta-analysis, Journal of Allergy and Clinical Immunology (2017), doi: /j.jaci This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

2 1 1 2 TITLE PAGE Diagnostic accuracy of fractional exhaled nitric oxide in predicting cough variant asthma and eosinophilic bronchitis in adult patients with chronic cough: A systematic review and meta-analysis Woo-Jung Song, MD 1, Hyun Jung Kim, MPH, PhD 2, Ji-Su Shim, MD 1, Ha-Kyeong Won, MD 1, Sung-Yoon Kang, MD 1, Kyoung-Hee Sohn, MD 1, Byung-Keun Kim, MD 1,3, Eun-Jung Jo, MD 4, Min-Hye Kim, MD, PhD 5, Sang-Heon Kim, MD, PhD 6, Heung-Woo Park, MD, PhD 1, Sun-Sin Kim, MD, PhD 1, Yoon-Seok Chang, MD, PhD 1,3, Alyn H Morice, MD, FRCP 7, Byung-Jae Lee, MD, PhD 8, Sang-Heon Cho, MD, PhD 1 1 Division of Allergy and Clinical Immunology, Department of Internal Medicine, Seoul National University College of Medicine, Seoul, Korea 2 Department of Preventive Medicine, Korea University College of Medicine, Seoul, Korea 3 Division of Allergy and Clinical Immunology, Department of Internal Medicine, Seoul National University Bundang Hospital, Seongnam, Korea 4 Department of Internal Medicine, Pusan National University College of Medicine, Busan, Korea 5 Department of Internal Medicine, Ewha Woman s University School of Medicine, Seoul, Korea 6 Department of Internal Medicine, Hanyang University College of Medicine, Seoul, Korea 7 Centre for Cardiovascular and Metabolic Research, Hull York Medical School, Castle Hill Hospital, University of Hull, Cottingham, East Yorkshire, United Kingdom Division of Allergy, Department of Medicine, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea

3 Co-Correspondence to: Byung-Jae Lee, MD, PhD, Division of Allergy, Department of Medicine, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, Korea; Tel: ; Fax: ; Sang-Heon Cho, MD, PhD, Division of Allergy and Clinical Immunology, Department of Internal Medicine, Seoul National University College of Medicine, Seoul, Korea; Tel: ; Fax: ; Word count: 3,696 (excluding title page, abstract, reference, figure and tables) The number of table: 1 The number of figure: 5 Source of funding: This study was supported by a grant no from the SNUH Research Fund. Conflict of interest statement: None to declare

4 3 40 Abstract 41 Background: Individual studies suggested the utility of fractional exhaled nitric oxide (FeNO) in detecting cough variant asthma (CVA) and eosinophilic bronchitis (EB) in chronic cough patients. Objective: To obtain summary estimates of diagnostic test accuracy of FeNO in predicting CVA and/or EB in adult patients with chronic cough Methods: Electronic databases were searched for studies published until January 2016, without language restriction. Cross-sectional studies which reported the diagnostic accuracy of FeNO for detecting CVA or EB were included. Risk of bias was assessed with QUADAS-2. Random effects meta-analyses were performed to obtain summary estimates of the diagnostic accuracy of FeNO. Results: A total of 15 studies involving 2,187 adult patients with chronic cough were identified. FeNO had a moderate diagnostic accuracy in predicting CVA in patients with chronic cough, showing the summary area under curve (AUC) to be 0.87 (95% CI: ). Specificity was higher and more consistent than sensitivity (0.85 [95% CI: ] and 0.72 [95% CI: ], respectively). However, in non-asthmatic chronic cough population, the diagnostic accuracy to predict EB was found to be relatively lower (the summary AUC 0.81 [95% CI: ]) and specificity was inconsistent. Conclusions: The present meta-analyses indicated the diagnostic potential of FeNO as a rule in 59 test for detecting CVA in adult patients with chronic cough. However, FeNO may not be useful

5 to predict EB in non-asthmatic chronic cough. These findings warrant further studies to validate the roles of FeNO in clinical practice of chronic cough patients Word count: 246 Key Messages Fractional exhaled nitric oxide (FeNO) has moderate diagnostic accuracy for predicting cough variant asthma (CVA) or eosinophilic bronchitis in patients with chronic cough. Optimal diagnostic points obtained from individual studies showed that specificity is higher and more consistent than sensitivity in predicting CVA. Capsule summary Fractional exhaled nitric oxide has a diagnostic potential as a rule in test for cough variant asthma in patients with chronic cough. Keywords: chronic cough; fractional exhaled nitric oxide; asthma; eosinophilic bronchitis; systematic review

6 5 77 Abbreviations 78 Th2, Type 2 helper T cell CVA, cough variant asthma EB, eosinophilic bronchitis NO, nitric oxide FeNO, fractional exhaled nitric oxide ATS, American Thoracic Society RQ, research question ROC, receiver operating characteristic AUC, area under the curve QUADAS-2, Quality Assessment of Diagnostic Accuracy Studies 2 HSROC, hierarchical summary receiver operating characteristic AHR, airway hyper-responsiveness 95% CI, 95% confidence interval

7 Introduction Chronic cough is a common medical condition with a significant impact on quality of life. 1, Several intrinsic and extrinsic factors may contribute to hypersensitivity in the cough reflex, and Type 2 helper T cell (Th2)-mediated airway inflammation is one of the major triggers of this condition. 3,4 The diagnostic approaches for cough variant asthma (CVA) and eosinophilic bronchitis (EB), two common conditions with Th2 inflammation, have been an integral part of the clinical guidelines for chronic cough. 5,6 Notably, these are particular conditions where objective diagnostic tests are preferred to empirical treatment in suspected patients during the early stages of their diagnostic work-up. 5,6 However, the tests required for diagnosing CVA and EB, such as bronchial challenge tests and induced sputum analyses, are technically demanding and require specialised instruments and personnel. 7-9 Thus, their use is restricted to specialist centres. Since the discovery of the biological roles of endogenous nitric oxide (NO) and the standardisation of simple analysers of exhaled NO in the 1990s 10,11, fractional exhaled nitric oxide (FeNO) has been suggested as a potential biomarker for Th2 inflammation. 12 The great advantage of measuring FeNO is that it only requires a simple, rapid, and non-invasive test 13, potentially enabling the test to be widely used in clinical practice. International clinical practice guidelines by the American Thoracic Society (ATS) on FeNO measurement and interpretation suggest its utility in the diagnosis and monitoring of asthma, and also its potential value for chronic obstructive pulmonary disease and pulmonary hypertension. 13 However, its diagnostic utility may vary with target population, and no specific recommendation has been made for 112 patients with chronic cough. 13

8 Considering convenience and safety, FeNO may be integrated into the diagnostic pathway for chronic cough patients. In the literature, individual studies have suggested diagnostic value in detecting CVA or non-asthmatic EB. 14,15 However, the diagnostic properties of FeNO measurements for CVA and EB have not been systematically reviewed within the clinical context of chronic cough. Here, we report a systematic review, summarising the current evidence. We also conducted meta-analyses to obtain summary estimates of diagnostic test accuracy for FeNO in predicting CVA and/or EB in adult patients with chronic cough.

9 Methods Search strategy We conducted a systematic literature search to identify studies that had relevant information for the following three research questions (RQs): What is the diagnostic accuracy of FeNO (index test) for CVA and/or EB (reference standard) in patients with chronic cough (target condition)? RQ1: FeNO for CVA in chronic cough patients. RQ2: FeNO for either CVA or EB in chronic cough patients RQ3: FeNO for EB in non-asthmatic chronic cough patients. We searched the Pubmed, Embase, Web of Science, and Scopus databases according to the recommendation of the PRISMA statement. 16 The search terms were cough AND nitric oxide in [all fields] that correspond to the target condition (chronic cough) and index test (FeNO), which are the two main components of our RQs. The search encompassed studies published up to January 2016, with no language restriction. We used this search strategy because we previously found that the number of publications was not excessive. 17 Additional searches were performed using Google Scholar and all cross-referenced articles. The eligibility criteria are provided in the Methods section of Online Repository. The study was registered as PROSPERO CRD (date of registration: 15 March 2016). Data extraction Data were independently extracted by two authors. We constructed 2 2 tables describing the index test results for each reported threshold level of FeNO. If 2 2 tables were not described in

10 the original publications, we reconstructed them from summary statistics, or calculated them from the information provided by the corresponding authors. The data extraction methods detailed in Online Repository. Quality assessment The methodological quality of the studies included was independently assessed by two authors, using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool, which consists of four domains. 18 The details of quality assessment are provided in Online Repository. Statistical analyses From the constructed 2 2 table, we calculated estimates of sensitivity and specificity, as well as 95% confidence intervals (CIs) for the individual study. The data were summarised as paired forest plots. Summary ROC curves were created to examine the overall diagnostic properties of FeNO in each RQ. We used a random effects hierarchical summary ROC (HSROC) model to adjust for between-study variability in cut-off levels. 19 For the HSROC curve, the individual and summary points for sensitivity and specificity, as well as the 95% confidence region around the summary points and the 95% prediction region were plotted. The 95% prediction region shows the amount of statistical heterogeneity between studies and predicts the sensitivity and specificity in a future study. Heterogeneity was assessed first by the visual assessment of forest and HSROC plots Threshold effects were determined to be likely if the Spearman correlation coefficient between sensitivity and the false positive rate was Cook's distances were used to identify outliers

11 that influenced the summary estimates. The I 2 statistic was calculated to quantify the unobserved heterogeneity, which describes the percent of total variation across studies that is attributable to heterogeneity instead of chance. 21 The I 2 statistic, however, is limited in diagnostic test reviews, as it does not account for the variation due to the threshold effect. 22 To further explore the source of heterogeneity, sensitivity analyses and multivariate meta-regression analyses were performed. Meta-regression was used to evaluate the statistical significance of differences in sensitivity and specificity by study characteristics. In multivariate analyses, covariates were selected if they were considered to contribute to the heterogeneity in univariate meta-regression analyses (p<0.05). Subgroup analyses were performed on multiple pre-determined covariates, including cough duration, study design, and outcome definitions. All of the statistical analyses were performed using the RevMan software (ver. 5.3; Cochrane Collaboration, Oxford, UK) and the meta-analysis modules of Stata (ver. 14.1; Stata Corp., College Station, TX, USA) (metandi and midas).

12 Results Study characteristics and methodological quality In total, 15 studies involving 2,187 adult patients with chronic cough were identified to respond to one of the three RQs (Fig. 1). Baseline characteristics of the 15 included studies are summarised in Table S1 and the Results section of Online Repository. Fourteen studies were published as full papers 14,15,23-34, and one was a conference abstract. 35 Most of the studies were conducted in Asia (10/15; 8 in China, 1 in Korea, 1 in Japan), and five were performed in Europe and Canada. Selection criteria from each included study are summarised in Table S2 and Online Repository. The studies excluded at the final stage are detailed in Table S3 and Online Repository. The most common criterion for defining CVA was the combination of (1) cough (as the sole or predominant symptom); (2) demonstration of airway hyper-responsiveness (AHR) and/or variable airway obstruction; and (3) a good treatment response to asthma therapy (10 studies) 24-27,30-35, which followed current international or national clinical guidelines 5,6,36 (thus termed the guideline-based diagnosis of CVA). The combination of (1) cough and (2) demonstration of AHR and/or variable airway obstruction was adapted in five studies 14,15,23,28,29 (the presumed diagnosis of CVA). The latter was mostly used in the studies published before EB was defined as induced sputum eosinophilia ( 3% in 2 studies 15,29 ), or as the combination of (1) a good treatment response to corticosteroids and (2) induced sputum eosinophilia ( 2.5% in three studies 24,32,33, 2% in one study 27, and 3% in one study 25 ); here we allowed these variations in the criteria for sputum eosinophilia, as each value is used in international or national clinical 201 guidelines. 5,6,36,37

13 The quality of the studies included was assessed by QUADAS-2 (detailed in Fig. S1 and Online Repository) RQ1: Diagnostic utility of FeNO for cough variant asthma in chronic cough In total, 13 studies (2,019 patients) provided diagnostic information of FeNO for CVA in chronic cough patients. 14,15,23,24,26,28-35 Because the definition for chronic cough ( 8 vs. 3 weeks) was a potential source of heterogeneity (Online Repository Table S4), we divided the analyses according to the definition. First, we analysed 10 studies (1,793 patients) that defined chronic cough as cough 8 weeks. 23,24,26,28,30-35 The findings on diagnostic accuracy are shown as a forest plot (Fig. 2A). Optimal cut-off levels ranged from 15.9 to 55.0 ppb, and were between 30 and 40 ppb in 8 of 10 studies. Figure 3A shows the HSROC curve for the primary meta-analysis. The visual distribution of 10 optimal diagnostic points obtained from the individual studies indicated that specificity was higher and more consistent than sensitivity. Sensitivity and specificity were 0.72 (95% CI: ; 95% prediction: ; I 2 =91.9%) and 0.85 (95% CI: ; 95% prediction: ; I 2 =47.6%), respectively. Spearman s correlation tests for sensitivity and false positive rate suggested that the heterogeneity was likely to be due to non-threshold effects (r=-0.309, p=0.385). The summary area under curve (AUC) was 0.87 (95% CI: ) (Table 1). Two studies had notably lower sensitivities 26,34 and Cook s distance analyses revealed that they 26,34 were outliers showing a relatively higher proportion of CVA (48.8% and 62.2% respectively) 26,34 (Online Repository Fig. S2). In linear regression analyses, the proportion of CVA was correlated with sensitivity (r 2 =0.321, p=0.087). These findings led us to speculate if methodological difference (such as inclusion criteria) could have influenced the sensitivity. The

14 study by Ni 26 had a retrospective design and a presumed diagnosis of CVA, while the study by Zhu 34 did not include normal chest X-rays in their selection criteria (Online Repository Table S and S2), suggesting that CVA diagnoses could be confounded in those studies. Thus, those studies were excluded from the primary meta-analysis to examine the effects. The resultant sensitivity and specificity were 0.75 (95% CI: ) and 0.87 (95% CI: ), respectively. To further explore the potential sources of heterogeneity, sensitivity analyses were performed (Online Repository Table S5). Several covariates, including studies from Asian region, a non-english publication, a retrospective design, utilisation of electrochemical FeNO analysers, utilisation of a presumed diagnosis of CVA, or the lack of normal chest X-rays in selection criteria, led to lower sensitivities. In multivariate meta-regression analyses, the CVA diagnostic criteria and the type of FeNO analyser were significant sources of heterogeneity (Online Repository Table S6). Details on meta-regression analyses are provided in Online Repository. Subgroup analyses were performed on the study design and CVA diagnostic criteria. In the subgroup analysis of prospective studies (n = 7; Online Repository Fig. S3A) 23,28,30,32-35, the sensitivity, specificity, and summary AUC were 0.73 (95% CI: ; 95% prediction: ; I 2 =84.0%), 0.86 (95% CI: ; 95% prediction: ; I 2 =25.3%), and 0.88 (95% CI: ), respectively (Fig. 3B). In the subgroup analysis of studies with guideline-based diagnostic criteria for CVA (n = 7; Online Repository Fig. S3B) 24,30-35, the sensitivity, specificity, and summary AUC were 0.71 (95% CI: ; 95% prediction: ; I 2 =65.3%), 0.88 (95% CI: ; 95% prediction: ; I 2 =43.1%) and 0.89 (95% CI: ), 246 respectively (Fig. 3C).

15 Additional subgroup analyses were performed post hoc on studies that met the diagnostic recommendations in current guidelines 5,6, and those in which normal chest X-rays and guideline-based diagnostic criteria for CVA were used; six studies met these criteria. 24,30-33,35 Optimal cut-off levels ranged from 15.9 to 55 ppb (median 36.7 ppb), and 4 studies ranged within ppb (Online Repository Fig. S3C). 24,30,32,33 The pooled sensitivity and specificity were 0.75 (95% CI: ; 95% prediction: ; I 2 =0.0) and 0.87 (95% CI: ; 95% prediction: ; I 2 =30.1%) respectively, suggesting that a strict indication of FeNO within the diagnostic pathways could result in better sensitivity (Fig. 3D). There were three studies (226 patients) with cough 3 weeks (Online Repository Fig. S4). 14,15,29 This pool was enriched with unpublished data from one study 15 that were obtained by contacting the corresponding author. However, the minimum required number (n = 4) for performing a meta-analysis was not met. The optimal cut-off level ranged from 30.0 to 42.5 ppb. Sensitivity and specificity were in the ranges and , respectively. RQ2: Diagnostic utility of FeNO for either cough variant asthma or eosinophilic bronchitis in chronic cough Four studies of 529 patients were analysed (Fig. 2B). 15,25,27,32 This analysis was enriched with unpublished data from the study by Oh et al. 15 Optimal cut-off levels ranged from 31.5 to 42.5 ppb. Sensitivity was 0.72 (95% CI: ; 95% prediction: ; I 2 =77.7%); one study 25 showed a particularly high sensitivity value of 0.96, whereas others showed lower values (range: ). Specificity was 0.85 (95% CI: ; 95% prediction: ; I 2 =7.5%). The HSROC curve analyses estimated the summary AUC to be 0.89 (95% CI: ; Fig. 4). Heterogeneity, particularly with regard to sensitivity, was suggested by the HSROC

16 curve appearance; however, further analysis was not possible due to a small number of included studies RQ3: Diagnostic utility of FeNO for eosinophilic bronchitis in non-asthmatic chronic cough Four studies of 390 non-asthmatic chronic cough patients were analysed (Fig. 2C). 15,24,32,33 Optimal cut-off levels ranged from 22.5 to 31.7 ppb. Sensitivity was estimated as 0.72 (95% CI: ; 95% prediction: ; I 2 =3.6%), and specificity was estimated as 0.83 (95% CI: ; 95% prediction: ; I =61.8%). In the HSROC curve analysis, the summary AUC was calculated as 0.81 (95% CI: ; Fig. 5). A threshold effect was suggested by the appearance of the HSROC curve. However, further analysis for heterogeneity was not possible due to a small number of included studies.

17 Discussion FeNO can reflect Th2 airway inflammation, which may help define the Th2 endotype of chronic cough. Here, we identified the diagnostic potential of FeNO in predicting two major Th2 inflammatory conditions underlying chronic cough: CVA and EB. We found that FeNO measurements had moderate diagnostic accuracy in predicting CVA in adult patients with chronic cough. The estimated summary AUCs ranged from 0.81 to 0.89, according to the RQ, or in subgroup analyses. Sensitivities were highly variable across studies, ranging from 0.41 to 0.93; specificities were more consistent, ranging from 0.77 to A major outcome of the present study is that the performance characteristics and potential applicability of FeNO tests were reviewed systematically for the first time in the particular population of chronic cough. The moderate diagnostic accuracy demonstrated here indicated that the FeNO test alone is not sufficient to diagnose CVA in chronic cough patients (RQ1). However, a moderate diagnostic accuracy does not necessarily indicate a lack of utility; our meta-analyses suggested the direction of application of FeNO tests in that it showed a high diagnostic potential for consistency and accuracy as a rule-in test for CVA, rather than a rule-out test. Subgroup analyses restricted to the studies with prospective design or the guideline-based diagnostic criteria for CVA support our conclusion that high specificity is a potential strength of FeNO in the diagnosis of CVA. In turn, FeNO may not replace direct challenge tests such as the methacholine inhalation test, because the latter has the advantage of high sensitivity, and is useful to rule out current asthma (at a cut-off level of PC20 >16 mg/ml). 38 We observed that the sensitivities of FeNO were lower in two RQ1 studies. 26,34 Both studies had methodological concerns, including having a retrospective design or lacking chest X- rays in the selection criteria. Such limitations may have reduced the sensitivity of FeNO;

18 however, the summary sensitivity estimates remained at or below 0.80 in our sensitivity analyses excluding those two studies, or in other subgroup analyses. Sensitivity also varied with several conditions (Online Repository Table S5). We speculate that the variable FeNO sensitivity might be due to limitations in the diagnostic criteria for CVA. 5,6 Spirometry is frequently normal in chronic cough patients, and thus current clinical guidelines recommend methacholine inhalation challenge tests in the diagnosis of CVA. 5,6 Negative AHR can help to exclude the diagnosis of CVA, but positive AHR is just suggestive of, but not specific to asthma. 38 Also, determination of a positive treatment response (without placebo control) may not preclude spontaneous resolution of cough. In this regard, sensitivity is less likely to be a strength of FeNO in real-world settings. Performance characteristics were similarly demonstrated in the pooled analyses for FeNO in predicting either CVA or EB (RQ2). The pooled sensitivity and specificity were 0.73 (95% CI: ) and 0.89 (95% CI: ), respectively. Indeed, this topic has more clinical relevance, because EB is another clinical condition with good corticosteroid responsiveness 5,6, and thus evaluation for CVA alone has limited clinical value. However, only four studies were identified that were relevant to this topic. Thus, further studies are needed to validate the diagnostic role of FeNO for this purpose. Based on RQ1 analyses, we hypothesize that the performance characteristics observed in RQ2 were largely due to CVA predictability. Indeed, among 15 studies, there was a considerable proportion about CVA, comprising 36.8% (interquartile range %) of the study population. The optimal cut-off levels to predict CVA and/or EB were variable (RQ1 and RQ2), although they ranged within ppb in most studies (Figs. 3A, 4A). The variability in optimal cut-off levels may be due to the clinical heterogeneity between studies, such as population characteristics, selection criteria, FeNO measurement devices, or diagnostic procedures for target

19 conditions. In this study, we adjusted for threshold effects using random-effects model HSROC analyses. 19 However, due to the limited number of included studies, heterogeneity could not be fully assessed. Considering the performance characteristics and diagnostic potential of FeNO, we suggest further studies to determine diagnostic performance at relatively high cut-off levels (e.g., 30, 35, 40, or 50 ppb) or at a fixed specificity value (e.g., 0.95). Our second major finding was that the FeNO test may not be precise enough to predict EB in non-asthmatic chronic cough patients (RQ3). Our HSROC curve analyses suggested that neither sensitivity nor specificity was a potential advantage of FeNO in this clinical setting; the pooled sensitivity and specificity were 0.72 (95% CI: ) and 0.83 (95% CI: ), respectively. Because our findings were based on a small number of studies with some variations in the diagnostic procedure (n = 4), further validation is needed in larger populations. However, it is notable that a recent study of asthma patients 39 indicated less utility of FeNO for predicting airway eosinophilia. In the recent meta-analysis by Korevaar et al. 39, FeNO showed moderate diagnostic accuracy, but low sensitivity and specificity for detecting sputum eosinophilia ( 3% of induced sputum) in adult patients with asthma; the summary AUC was 0.74 (95% CI: ), and the pooled sensitivity and specificity were 0.66 (95% CI: ) and 0.76 (95% CI: ), respectively. The regulatory mechanism of FeNO levels may be distinct from that of sputum eosinophilia 12, as suggested in clinical trials of mepolizumab (anti-il-5 monoclonal antibody) in severe asthma patients. 40,41 These findings indicate that FeNO may not fully replace induced sputum analyses. Also, we suggest that these findings collectively challenge the statement in the current ATS guidelines 13, which strongly recommend the use of 349 FeNO in the diagnosis of eosinophilic airway inflammation.

20 The clinical utility of FeNO, however, should not be dismissed in the non-asthmatic cough population. The induced sputum test has become the best proxy to define lower airway eosinophilia, but there have been concerns about its feasibility and reproducibility. 42 Considering the historical background for developing the clinical concept of non-asthmatic EB (which was to identify corticosteroid-responsive patients among the non-asthmatic cough population) 43, additional studies are needed to investigate the utility of FeNO in predicting ICS responsiveness rather than predicting sputum eosinophilia. To the best of our knowledge, no studies have used validated cough measurement tools to examine such utility in patients with non-asthmatic chronic cough. 17 Several major limitations need to be considered in interpreting our findings. First, heterogeneity could not be fully investigated due to the limited number of studies included. Several factors, such as demographics, selection criteria, clinical settings, FeNO measurement device, and definitions of target conditions, may have contributed to the observed heterogeneity. Second, the majority of studies included (10/15) were conducted in Asia, and seven were published in Chinese language journals. Thus, issues may be raised about generalisability. We could not fully examine regional differences, because the number of studies from Western populations was limited. However, specificity was comparable between Asian and Western regions (Online Repository Table S5). Third, publication bias could not be determined, because of the lack of techniques for detecting such biases in diagnostic test accuracy reviews. 44 However, publication bias was minimised by including non-english literature, conference abstracts and unpublished data. Fourth, current diagnostic criteria for CVA may not be precise, as 1) it is often based on positive AHR and may not preclude post-vial cough 45 and 2) premises good clinical response to asthma therapy and may not encompass refractory asthma. However, CVA is

21 generally considered to respond well to asthma treatment, and any heterogeneity of treatment response in patients with CVA has not been addressed yet. Collectively, the present findings on CVA need to be interpreted cautiously and within the context of current clinical guidelines for chronic cough. 5,6 Finally, the possibility of overestimated diagnostic accuracy may be present; we tried to minimise this risk by excluding case-control studies and including unpublished data and non-english literature in the process of study selection. However, all of our included studies were derivation studies, which could lead to overoptimistic estimations of sensitivity and specificity. 18 However, we argue that this could be a paradoxical strength of the study, because the diagnostic potential for its high specificity has not been reported in the chronic cough population before. These findings could help to develop a further validation study at a fixed cutoff (or specificity) level. These efforts could finally lead to the development of specific recommendations on the use of FeNO in clinical guidelines for chronic cough patients. In conclusion, the present systematic review and meta-analysis demonstrated overall moderate diagnostic accuracy of FeNO in predicting CVA and/or EB in adult patients with chronic cough. Pooled analyses indicated the diagnostic potential of FeNO as a rule in test for CVA. The number of studies on EB was small, but available evidence suggested that FeNO was less useful to predict EB in non-asthmatic chronic cough patients. Further studies are necessary to validate the best ways to use the FeNO test in chronic cough patients.

22 Acknowledgement We would like to thank Prof Heikki Koskela (Kuopio University, Finland) for his advice on data analyses. Also we would like to thank Prof Krzysztof Kowal (Medical University of Bialystok, Poland), Prof Mauro Maniscalco (Hospital S. Maria della Pieta, Italy) and Dr Stefano Pizzimenti (University of Torino, Italy) for providing the information on their study design. Finally, we would like to thank Prof Mi-Jung Oh (Bundang Jaeseng Hospital, Korea) and Prof Dong-Chull Choi (Sungkyunkwan University, Korea) for providing the original data.

23 Reference 1. Song WJ, Chang YS, Faruqi S, Kim JY, Kang MG, Kim S, et al. The global epidemiology of chronic cough in adults: a systematic review and meta-analysis. Eur Respir J 2015;45: French CL, Irwin RS, Curley FJ, Krikorian CJ. Impact of chronic cough on quality of life. Arch Intern Med 1998;158: Song WJ, Chang YS. Cough hypersensitivity as a neuro-immune interaction. Clin Transl Allergy 2015;5: Morice AH, Millqvist E, Belvisi MG, Bieksiene K, Birring SS, Chung KF, et al. Expert opinion on the cough hypersensitivity syndrome in respiratory medicine. Eur Respir J 2014;44: Morice A. The diagnosis and management of chronic cough. Eur Respir J 2004;24: Irwin RS, Baumann MH, Bolser DC, Boulet L-P, Braman SS, Brightling CE, et al. Diagnosis and management of cough executive summary: ACCP evidence-based clinical practice guidelines. Chest 2006;129:1S-23S. 7. Guidelines for methacholine and exercise challenge testing Am J Respir Crit Care Med 1999; Hamid Q, Kelly M, Linden M, Louis R, Pizzichini M, Pizzichini E, et al. Methods of sputum processing for cell counts, immunocytochemistry and in situ hybridisation. Eur Respir J 2002;20:19s-23s Paggiaro P, Chanez P, Holz O, Ind P, Djukanović R, Maestrelli P, et al. Sputum induction. Eur Respir J 2002;37:3s.

24 Barnes PJ, Belvisi M. Nitric oxide and lung disease. Thorax 1993;48: ATS/ERS recommendations for standardized procedures for the online and offline measurement of exhaled lower respiratory nitric oxide and nasal nitric oxide, Am J Respir Crit Care Med 2005;171: Alving K, Malinovschi A. Basic aspects of exhaled nitric oxide. Eur Respir Mon 2010; Dweik RA, Boggs PB, Erzurum SC, Irvin CG, Leigh MW, Lundberg JO, et al. An official ATS clinical practice guideline: interpretation of exhaled nitric oxide levels (FENO) for clinical applications. Am J Respir Crit Care Med Chatkin JM, Ansarin K, Silkoff PE, McClean P, Gutierrez C, Zamel N, et al. Exhaled nitric oxide as a noninvasive assessment of chronic cough. Am J Respir Crit Care Med 1999;159: Oh MJ, Lee JY, Lee BJ, Choi DC. Exhaled nitric oxide measurement is useful for the exclusion of nonasthmatic eosinophilic bronchitis in patients with chronic cough. Chest 2008;134: Moher D, Liberati A, Tetzlaff J, Altman DG, Group P. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. BMJ 2009;339:b Song WJ, Won HK, Moon SD, Chung SJ, Kang SY, Sohn KH, et al. Could fractional exhaled nitric oxide test be useful in predicting inhaled corticosteroid responsiveness in chronic cough? A systematic review. J Allergy Clin Immunol Pract 2016:In press. 18. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. 443 Ann Intern Med 2011;155:

25 Macaskill P, Gatsonis C, Deeks J, Harbord R, Takwoingi Y. Cochrane handbook for systematic reviews of diagnostic test accuracy. Version London: The Cochrane Collaboration Devillé WL, Buntinx F, Bouter LM, Montori VM, De Vet HC, Van der Windt DA, et al. Conducting systematic reviews of diagnostic studies: didactic guidelines. BMC Med Res Methodol 2002;2: Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med 2002;21: Campbell JM, Klugar M, Ding S, Carmody DP, Hakonsen SJ, Jadotte YT, et al. Diagnostic test accuracy: methods for systematic review and meta-analysis. Int J Evid Based Healthc 2015;13: Kowal K, Bodzenta-Lukaszyk A, Zukowski S. Exhaled nitric oxide in evaluation of young adults with chronic cough. J Asthma 2009;46: Lin G. The clinical diagnosis value of exhaled nitric oxide test in chronic cough. Chin Modern Med 2013;20: Maniscalco M, Faraone S, Sofia M, Molino A, Vatrella A, Zedda A. Extended analysis of exhaled and nasal nitric oxide for the evaluation of chronic cough. Respir Med 2015;109: Ni J, Cheng Q, Feng Y, Cao B, Cheng L, Wan H. Exhaled nitric oxide combined with relevant factors in the diagnosis of cough variant asthma. J Diagn Concepts Pract 2014;13: Pacheco A, Faro V, Cobeta I, Royuela A, Molyneux I, Morice AH. Gastro-oesophageal reflux, eosinophilic airway inflammation and chronic cough. Respirology 2011;16:994-9.

26 Qiu JP, Jin XY, Shen HY. Application of fractional exhaled nitric oxide and impulse oscillometry in patients with chronic cough. Int J Respir 2012;32: Sato S, Saito J, Sato Y, Ishii T, Xintao W, Tanino Y, et al. Clinical usefulness of fractional exhaled nitric oxide for diagnosing prolonged cough. Respir Med 2008;102: Shen X, Chen C, Zhou N, Huang J, Xiuqin. Z. Feasibility study of fractional exhaled nitric oxide in the diagnosis of cough variant asthma. Int J Respir 2015;35: Ye L, Gong Y, Yueya. T. The etiological diagnosis value of fractional exhaled nitric oxide test in patients with chronic cough. J Clin Intern Med 2010;27: Yi F, Chen R, Luo W, Xu D, Han L, Liu B, et al. Validity of Fractional Exhaled Nitric Oxide in Diagnosis of Corticosteroids Responsive Cough. Chest Zhang YM, Lin JT, Su N, Chen X, Liu GL, Yu HX, et al. Values of fractional exhaled nitric oxide in the diagnosis of chronic cough. Natl Med J China 2011;91: Zhu N, He J, Chen XD. Diagnostic value of fractional exhaled nitric oxide in the diagnosis of cough variant asthma. J Clin Pulm Med 2014;19: Pizzimenti S, Heffler E, Piccioni P, Bugiani M, Migliore E, Guida G, et al. Usefulness of exhaled nitric oxide (FeNO) measured by a portable analyzer to diagnose Cough Variant asthma in a clinical setting of chronic cough. Allergy 2009;64: The Chinese national guidelines on diagnosis and management of cough (December 2010). Chin Med J (Engl) 2011;124: Reddel HK, Taylor DR, Bateman ED, Boulet L-P, Boushey HA, Busse WW, et al. An 488 official American Thoracic Society/European Respiratory Society statement: asthma

27 control and exacerbations: standardizing endpoints for clinical asthma trials and clinical practice. Am J Respir Crit Care Med 2009;180: Cockcroft DW. Direct challenge tests: Airway hyperresponsiveness in asthma: its measurement and clinical significance. Chest 2010;138:18S-24S. 39. Korevaar DA, Westerhof GA, Wang J, Cohen JF, Spijker R, Sterk PJ, 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: Flood-Page P, Swenson C, Faiferman I, Matthews J, Williams M, Brannick L, et al. A study to evaluate safety and efficacy of mepolizumab in patients with moderate persistent asthma. Am J Respir Crit Care Med 2007;176: Haldar P, Brightling CE, Hargadon B, Gupta S, Monteiro W, Sousa A, et al. Mepolizumab and exacerbations of refractory eosinophilic asthma. N Engl J Med 2009;360: Beghè B, Spanevello A, Fabbri LM. Eosinophilia in asthma: the easy way is not always the best. Lancet Respir Med 2015;3: Gibson P, Denburg J, Dolovich J, Ramsdale E, Hargreave F. Chronic cough: eosinophilic bronchitis without asthma. Lancet 1989;333: Leeflang MM. Systematic reviews and meta-analyses of diagnostic test accuracy. Clin Microbiol Infect 2014;20: Dicpinigaitis PV. Chronic cough due to asthma: ACCP evidence-based clinical practice guidelines. Chest 2006;129:75S-9S. 510

28 27 Table 1. Summary of meta-analyses by research question Number of studies Sensitivity (95% CI) Specificity (95% CI) Positive likelihood ratio (95% CI) RQ1. FeNO to predict cough variant asthma (in cough 8 weeks) Negative likelihood ratio Summary AUC (95% CI) (95% CI) CI) Diagnostic I 2 (95% CI) OR (95% ( ) 0.85 ( ) 4.7 ( ) 0.33 ( ) 0.87 ( ) 14 (8 26) 94 (89 99) RQ2. FeNO to predict either cough variant asthma or eosinophilic bronchitis ( ) 0.89 ( ) 6.5 ( ) 0.31 ( ) 0.89 ( ) 21 (8 54) 44 (0 100) RQ3. FeNO to predict non-asthmatic eosinophilic bronchitis ( ) 0.83 ( ) 4.1 ( ) 0.34 ( ) 0.81 ( ) 12 (6 23) 0 (0 100) Abbreviations: AUC, area under the curve; 95% CI, 95% confidence interval; RQ, research question; FeNO, fractional exhaled nitric oxide

29 28 Figure legends Figure 1. PRISMA for study selection Figure 2. Forest plots on the diagnostic utility of FeNO for (A) predicting cough variant asthma in 10 studies of chronic cough patients (defined by cough 8 weeks), (B) predicting either cough variant asthma or eosinophilic bronchitis in four studies of chronic cough patients, and (C) predicting eosinophilic bronchitis in four studies of non-asthmatic chronic cough patients. Figure 3. Hierarchical summary receiver operating characteristics (HSROC) curve on the diagnostic utility of FeNO for (A) predicting cough variant asthma (CVA) in 10 studies of chronic cough patients (defined by cough 8 weeks). Subgroup analyses (B) in 7 studies with a prospective design, (C) in 7 studies with guideline-based diagnostic criteria for CVA, and (D) in 6 studies which meet the diagnostic recommendations in international guidelines 5,6 (using normal chest X-rays and guideline-based diagnostic criteria for CVA). Size of the circle is proportionate to the size of study. Summary point indicates the summary values for sensitivity and specificity. A 95% prediction region is the confidence region for a forecast of the true sensitivity and specificity in future studies. HSROC curve is truncated outside of the area for which data points exist. Figure 4. Hierarchical summary receiver operating characteristics curve on the diagnostic utility of FeNO for predicting either cough variant asthma or eosinophilic bronchitis in four studies of chronic cough patients.

30 29 Figure 5. Hierarchical summary receiver operating characteristics curve on the diagnostic utility of FeNO for predicting eosinophilic bronchitis in four studies of non-asthmatic chronic cough patients.

31

32

33

34

35

36

37

38

39

40

41 1 1 Online Repository Diagnostic accuracy of fractional exhaled nitric oxide in predicting cough variant asthma and eosinophilic bronchitis in adult patients with chronic cough: A systematic review and meta-analysis Woo-Jung Song, MD, Hyun Jung Kim, MPH, PhD, Ji-Su Shim, MD, Ha-Kyeong Won, MD, Sung-Yoon Kang, MD, Kyoung-Hee Sohn, MD, Byung-Keun Kim, MD, Eun-Jung Jo, MD, Min-Hye Kim, MD, PhD, Sang-Heon Kim, MD, PhD, Heung-Woo Park, MD, PhD, Sun-Sin Kim, MD, PhD, Yoon-Seok Chang, MD, PhD, Alyn H Morice, MD, FRCP, Byung-Jae Lee, MD, PhD, Sang-Heon Cho, MD, PhD

42 2 11 Methods Search strategy Eligibility criteria were as follows. We included cross-sectional studies of adult patients presenting with chronic cough as their sole or predominant symptom, who reported the diagnostic properties of FeNO corresponding to one of the three RQs. We excluded case-control studies (such as a comparison study of non-asthmatic controls and CVA patients) to reduce the risk of selection bias; case-control studies may overestimate diagnostic accuracy, whereas crosssectional studies of unselected samples could reflect actual practice settings. 1 We also excluded studies if FeNO results were used in making the initial diagnosis of CVA or EB (to reduce expectation bias 2 ), or if the studies included patients with other respiratory symptoms (such as dyspnoea or wheezing). Publications without original data, such as review articles, were also excluded. Abstract only publications (such as conference abstracts) were not excluded if they contained original data. In cases where full texts or summary estimates were not available, we contacted the corresponding authors by s to obtain the information required for the analyses. Data extraction Data were independently extracted by two authors. Discrepancies were resolved by consensus. For all of the included studies, data were extracted for study design, clinical settings, index tests (FeNO), and reference standards (CVA and/or EB) in each target population (chronic cough or non-asthmatic chronic cough). We constructed 2 2 tables describing the results. In a case where a study did not report the optimal cut-off value and diagnostic accuracy, we performed receiver operating characteristic (ROC) curve analysis using the original data provided by the

43 corresponding authors, and determined the optimal cut-off value based on the best Youden index (sensitivity + specificity 1) Quality assessment The methodological quality of included studies were independently assessed by two authors, using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool, which consists of four domains. 3 In the domain of patient selection, we determined the risk of bias as low for prospective studies, which specified that they randomly or consecutively recruited the participants; retrospective studies were classified as unclear risk of bias due to potential bias originating from study design. In addition, studies with an insufficient description of the selection criteria were classified as having an unclear risk of bias. In the reference standard domain, studies were classified as having an unclear risk of bias if the diagnosis of CVA was not based on objective measurements (such as methacholine or histamine bronchial challenge tests). In the domain of flow and timing, studies were classified as having an unclear risk if they did not include all of the original participants in the analyses.

44 4 50 Results Study characteristics and methodological quality Twelve studies had prospective designs, whereas three had retrospective designs. Chronic cough was defined as cough 8 weeks in 12 studies, whereas it was defined as cough 3 weeks in 3 studies. Selection criteria from each study are summarised in Table S2; a normal chest X-ray and no current smoking were the most common criteria. There were considerable between-study differences in the FeNO measurement devices (three kinds of chemiluminescence analysers in six studies 4-9 and three kinds of electrochemical analysers in eight studies ; and not described in one study 18 ). Eight studies 5-7,11,12,15,16,18 reported adherence to international guidelines in measuring FeNO 19, six 4,8-10,13,14 referred to previous literature or manufacturer s instructions, and one (abstract) 17 did not provide such information. Two full papers 20,21 that reported the diagnostic accuracy of FeNO for CVA were excluded at the final stage. They referred to FeNO results in making the initial diagnosis of CVA, and thus were subject to expectation bias 2 (Table S3). The quality of the studies included was assessed by QUADAS-2 (Fig. S1). In the patient selection domain, seven prospective studies 4-7,11,13,15 were classified as low risk of bias, because they stated that participants were randomly or consecutively recruited. Three retrospective studies 8,10,12 and four prospective studies that did not state consecutive recruitment 9,14,16,18 were classified as having an unclear risk of bias. Detailed criteria could not be extracted from one study only published in abstract form 17, so it was deemed to have an unclear risk of bias. No study used a case-control design. In the index test domain, all of the studies used derivation tests (which fitted the cut-off level post hoc), not validation tests (with pre-determined cut-off levels). Thus, all were classified as having an unclear risk of bias; a derivation test is more likely

45 to overestimate diagnostic accuracy. 3 In all of the studies, FeNO (the index test) was measured prior to making a diagnosis of CVA/EB (reference standard). In the reference standard domain, all of the studies were classified as having a low risk of bias. In all of the studies, CVA and EB were defined on the basis of objective test results and international criteria. In addition, none of the studies used knowledge of FeNO results in making the diagnoses of CVA and EB. In the domain of flow and timing, all of the studies had a low risk of bias. All of the patients received the same reference standard within each study, and all of the recruited subjects were included in the analysis. RQ1: Diagnostic utility of FeNO for cough variant asthma in chronic cough Meta-regression analysis In univariate meta-regression analyses, study region, design, FeNO device type, and CVA diagnostic criteria were found to be potential sources of heterogeneity (p<0.05). In multivariate meta-regression analyses, CVA diagnostic criteria and FeNO analyser type were suggested to be potentially important sources of heterogeneity (Table S6). Considering the between-study variability of FeNO measurement devices, an additional meta-regression analysis was performed without this covariate and revealed that the diagnostic criteria of CVA (guideline-based diagnosis vs. presumed diagnosis) was a significant source of heterogeneity (data not shown). Meanwhile, inclusion of the covariate of publication language (English vs. non-english) did not significantly affect the outcome (data also not shown). Lack of normal chest X-rays in the selection criteria prevented inclusion into the multivariate models, as the effects could not be calculated. 95

46 6 Table S1. Baseline characteristics of the 15 studies included Study [Ref] Design n Mean Chatkin age (yr) Female (%) Location Prospective % Toronto, Canada Oh Prospective % Seoul, Korea Sato Prospective % Fukushima, Japan Kowal Prospective NA Bialystok, Poland Recruitment setting Tertiary referral clinic and affiliated community respiratory clinics University hospital University hospital Referral asthma clinic Definition of chronic cough Cough 3 weeks Cough 3 weeks Cough 3 weeks Cough 8 weeks FeNO measurement Chemiluminescence analyzer; Sievers NOA 280 (Sievers, Boulder, CO, USA) Chemiluminescence analyzer; Sievers NOA 280 (Sievers, Boulder, CO, USA) Chemiluminescence analyzer (Kimoto, Osaka, Japan) Chemiluminescence analyzer; Sievers NOA 280 (Sievers, Prevalence and Boulder, CO, USA) diagnostic criteria of cough variant asthma (%) 21%; Positive methacholine provocation test or positive bronchodilator response 12%; Positive methacholine provocation test (PC20 8 mg/ml) 67.7%; Positive methacholine provocation test or reversible airflow limitation 33%; Positive histamine provocation test or significant diurnal changes in peak expiratory flow (PEF) Prevalence and diagnostic criteria of non-asthmatic eosinophilic bronchitis (%) ND 18%; (1) Negative methacholine AHR; (2) Induced sputum eosinophils 3% 11.3%; (1) Negative methacholine AHR; (2) Induced sputum eosinophils 3% ND

47 7 Pizzimenti 2009 (abstract only) 17 Prospective 156 NA 59% Torino, Italy Ye Retrospective % Shanghai, China Pacheco Prospective % Madrid, Spain Respiratory Cough 8 Electrochemical or significant improvement of forced expiratory volume in 1 second (FEV1) medicine outpatient clinic weeks analyzer NIOX- MINO (Aerocrine AB, Solna, Sweden) (1) Positive methacholine provocation test (PD20<800 mu) and (2) response to antiasthmatic therapy (inhaled corticosteroids) University Cough 8 Chemiluminescence 41.9%; hospital weeks analyzer; NIOX (1) Normal spirometry, (Aerocrine AB, a positive Solna, Sweden) methacholine provocation test or daytime variability rate Tertiary 9%; of PEF >20%; (2) response to antiasthmatic therapy (bronchodilators and glucocorticosteroids). 22 Cough 8 Electrochemical 28.4%; cough clinic weeks analyzer; NIOX- (1) Positive MINO (Aerocrine methacholine AB, Solna, Sweden) provocation test or bronchodilator ND ND 4.3%; (1) Negative methacholine AHR; (2) induced sputum eosinophils 2% or

48 8 Zhang Prospective % Beijing, China Qiu Prospective % Shanghai, China Lin Retrospective % Guangzhou, Referral clinic Cough 8 weeks Electrochemical analyzer; NIOX- MINO (Aerocrine AB, Solna, Sweden) China Hospital clinic Hospital clinic Cough 8 weeks Cough 8 weeks Electrochemical analyzer; SV-02 Nano Coulomb Nitric Oxide Analyzer (Sunvou Medical Electronics Co., Ltd., Wuxi, China) Electrochemical analyzer; Nano Coulomb Nitric Oxide Analyzer (Sunvou Medical response; (2) response to anti-asthmatic therapy %; (1) Normal spirometry, a positive methacholine provocation test or daytime variability rate of PEF >20%; (2) response to antiasthmatic therapy (bronchodilators and glucocorticosteroids) %; Positive methacholine provocation test or bronchodilator response test 44%; (1) Normal spirometry, a positive methacholine provocation test or bronchoalveolar lavage fluid eosinophils 4%; (3) response to corticosteroids 28.3%; (1) Normal spirometry without AHR; (2) Induced sputum eosinophils 2.5%; (3) response to corticosteroids. 22 ND 35%; (1) Normal spirometry without AHR; (2) Induced sputum

49 9 Ni Retrospective 297 NA 49% Shanghai, China Zhu Prospective % Shanghai, China Maniscalco Prospective % Naples, Italy Electronics Co., Ltd., Wuxi, China) daytime variability rate of PEF >20%; (2) response to antiasthmatic therapy (bronchodilators and glucocorticosteroids). 22 University Cough 8 Electrochemical 48.8%; hospital weeks analyzer; Nano Positive histamine University hospital Primary clinic Cough 8 weeks Cough 8 weeks Coulomb Nitric Oxide Analyzer (Sunvou Medical Electronics Co., Ltd., Wuxi, China) Chemiluminescence analyzer; NIOX (Aerocrine AB, Solna, Sweden) Electrochemical analyzer; Hypair FeNO system (Medisoft, Sorinnes, Belgium) provocation test 62.2%; (1) Normal spirometry, a positive histamine provocation test or daytime variability rate of PEF >20%; (2) response to antiasthmatic therapy (bronchodilators and glucocorticosteroids) %; (1) Positive methacholine provocation test; (2) response to anti- eosinophils 2.5%; (3) response to corticosteroids 22. ND ND 19.2%; (1) Normal spirometry without variable airflow obstruction or AHR;

50 10 Shen Prospective % Suzhou, China Yi Prospective % Guangzhou, China University Hospital University Hospital Cough 8 weeks Cough 8 weeks ND 46.5%; Electrochemical analyzer; NIOX- MINO (Aerocrine AB, Solna, Sweden) asthmatic therapy (bronchodilators and glucocorticosteroids). 24 (1) Normal spirometry, a positive histamine provocation test or daytime variability rate of PEF >20%; (2) response to antiasthmatic therapy (bronchodilators and glucocorticosteroids) %; (1) Normal spirometry, a positive methacholine provocation test or daytime variability rate of PEF >20%; (2) response to antiasthmatic therapy (bronchodilators and glucocorticosteroids). 22 (2) induced sputum eosinophils 3%; (3) resolution of cough by corticosteroid therapy. 24 ND 21.7%; (1) Normal spirometry without AHR; (2) Induced sputum eosinophils 2.5%; (3) response to corticosteroids. 22 FEV1, forced expiratory volume in 1 second; PC20, provocative concentration of methacholine causing a 20% fall in FEV1; PD20, provocative dose of methacholine causing a 20% fall in FEV1; PEF, peak expiratory flow; AHR, airway hyper-responsiveness; ND, not described

51 11

52 12 Table S2. Summary of the selection criteria for participants in 15 included studies Study [ref] Age Chest X-ray Lung function Smoking Medication Exclusion of other history history medical conditions Chatkin Adult (age Normal FEV1>80% of No current No codeine or No other significant criteria not predicted smoker specified) Oh Adult (age Normal FEV1 80% of No criteria not predicted specified) current smoker or exsmoker who other medication No cough corticosteroids within 6 weeks No anti-asthma medications within 2 months quitted within 1 year Sato years Normal No prior corticosteroid use Kowal years Normal Within normal Non-smoker No prior use of range any anti-asthma medication No angiotensin converting enzyme inhibitor (ACEi) medical conditions No wheezing or crackles on physical examination No history of treatment pulmonary disease for No history of any systemic disease History of respiratory tract infection No upper respiratory tract infection within 4 weeks No upper respiratory tract infection within 4 weeks Other No contraindication to methacholine challenge No contraindication to histamine challenge

53 13 Adult (age No codeine or other cough suppressant Pizzimenti 2009 Normal No current (Abstract only) 17 criteria not smoker specified) Ye years Normal No smoking history No inhaled corticosteroid within 4 weeks No bronchodilator use on the test day No allergy No history of chronic bronchitis, chronic obstructive pulmonary (COPD), bronchiectasis medication on the test day Pacheco Adults (age No current No COPD criteria not smoker specified) Zhang years Normal No smoking history No corticosteroids within 4 weeks Qiu years Normal No smoking No use of shortacting within 4 hours on the bronchodilator disease or No history of serious diseases No history of chronic bronchitis, COPD, or No upper respiratory tract infection within 4 weeks No respiratory tract infection within 2 weeks No upper respiratory tract infection within 4 No allergic rhinitis episodes within 4 weeks No allergic rhinitis episodes within 4 weeks No allergic rhinitis episodes within 4 weeks

54 14 test day within 4 hours on the test day No corticosteroids within 4 weeks bronchiectasis Lin Adults (age Normal No smoking No history of criteria not history specified) Ni years Normal No current smoker No corticosteroids within 4 weeks No use of shortacting bronchodilator and allergy chronic lung diseases and other systemic diseases No history of chronic COPD, bronchiectasis bronchitis, medication within 2 days Zhu Adults (age FEV1 70% of No long-term No history of No history of severe criteria not predicted smoker specified) corticosteroid use No longterm use of ACEi or heart and lung disease weeks No upper respiratory tract infection within 4 weeks No upper respiratory tract infection within 2 weeks No allergic rhinitis episodes within 4 weeks No pregnancy Maniscalco Adults (age criteria not Normal FEV1 80% of predicted No smoker current No current medication for No wheezing or crackles on physical

55 15 specified) cough examination Shen years Normal No current smoker Yi years Normal No current smoker or exsmoker who quitted within 6 months No corticosteroids within 4 weeks No use of shortacting bronchodilator on the test day No corticosteroids oral within 4 weeks No history of other respiratory systemic diseases or No upper respiratory tract infection within 2 weeks No upper respiratory tract infection within 8 weeks No allergic rhinitis episodes within 4 weeks

56 16 Table S3. Summary of 2 studies which were excluded in the final stage of selection Study [Ref] Design n Mean age (yr) Female (%) Location Zeng Prospective % Shanghai, China Wu Prospective % Shenzhen, China Recruitment setting Referral hospital University affiliated hospital Definition of chronic cough Cough 8 weeks Cough 8 weeks FeNO measurement Electrochemical analyzer; Nano Coulomb Nitric Oxide Analyzer SV- 02 (Sunvou Medical Electronics Co., Ltd., Wuxi, China) Electrochemical analyzer; Nano Coulomb Nitric Oxide Analyzer SVeNO-1 (Sunvou Medical Electronics Prevalence and diagnostic criteria of cough variant asthma (%) 45%; Initial diagnosis: FeNO 35 ppb Final diagnosis: (1) Patients with FeNO 35 ppb were treated with inhaled corticosteroid and bronchodilator for 1 month, and those with FeNO<35 ppb were given different targeted therapy. (2) Final diagnosis was made on the treatment outcome. 46%; Initial diagnosis: FeNO 35 ppb Final diagnosis: (1) Patients with FeNO 35 ppb were treated Prevalence and diagnostic criteria of non-asthmatic eosinophilic bronchitis (%) ND ND

57 17 ND, not described Co., Ltd., Wuxi, China) with inhaled corticosteroid and bronchodilator for 3 months, and those with FeNO<35 ppb were given different targeted therapy. (2) Final diagnosis was made on the treatment outcome.

58 18 Table S4. Univariate meta-regression analyses for possible source of heterogeneity in 13 studies reporting the diagnostic utility of FeNO for predicting cough variant asthma Parameter Category n Sensitivity p value Specificity p value Cough duration 8 weeks ( ) ( ) < weeks ( ) 0.82 ( )

59 19 Table S5. Sensitivity analyses for possible sources of heterogeneity in 10 studies reporting the diagnostic utility of FeNO for predicting cough variant asthma in patients with chronic cough ( 8 weeks) Parameter Category n Sensitivity (95% CI) Specificity (95% CI) Region Asian ( ) 0.84 ( ) Western ( ) 0.85 ( ) Publication language Non-English ( ) 0.85 ( ) English ( ) 0.85 ( ) Study design Prospective ( ) 0.84 ( ) Retrospective ( ) 0.86 ( ) FeNO analyser type * Electrochemical analyser ( ) 0.85 ( ) Chemiluminescence analyser ( ) 0.83 ( ) Diagnostic criteria of Guideline-based diagnosis (including the criteria of good ( ) 0.87 ( ) cough variant asthma treatment response to asthma therapy) Presumed diagnosis (not including the criteria of good treatment ( ) 0.81 ( ) response to asthma therapy) Normal chest X-ray as Described ( ) 0.84 ( ) selection criteria Not described ( ) 1.00 ( ) * One study did not provide the device information.

60 20 Table S6. Multivariate meta-regression analyses for possible sources of heterogeneity in 10 studies reporting the diagnostic utility of FeNO for predicting cough variant asthma in patients with chronic cough ( 8 weeks) Parameter Category Likelihood ratio test chi-squared p value I 2 Region Asian vs. Western % (0 100) Study design Prospective vs. retrospective % (0 100) FeNO analyser type Electrochemical vs. chemiluminescence analyser < % (32 100) Diagnostic criteria for cough variant asthma Guideline-based diagnosis (including the criteria of good treatment response to asthma therapy) vs. presumed diagnosis (not including the criteria of good treatment response to asthma therapy) % (68 100)

61 Fig. S1. QUADAS-2 quality assessment of the studies included (A) Graph showing the risk of bias and concerns regarding applicability: review of authors judgements in each domain, presented as percentages across the studies included. (B) Risk of bias and concerns regarding applicability summary: review of authors judgements in each domain for all studies included. (A)

62 22 8 (B) 9