Curbing the Cough: Multimodal Treatments for Neurogenic Cough: A Systematic Review and Meta-Analysis

Abstract

Objectives/Hypothesis:

Neurogenic cough affects 11% of Americans and causes significant detriment to quality of life. With the advent of novel therapies, the objective of this review is to determine how procedural therapies (e.g., superior laryngeal nerve block) compare to other established pharmacologic and non-pharmacologic treatments for neurogenic cough.

Methods:

With the assistance of a medical librarian, a systematic review was performed using PICOS (patients, interventions, comparator, outcome, study design) format: adults with neurogenic cough receiving any pharmacologic or non-pharmacologic treatment for neurogenic cough compared to adults with neurogenic cough receiving any other relevant interventions, or treated as single cohorts, assessed with cough-specific quality of life outcomes, in all study designs and case series with ≥ 10 cases. Case reports, review articles, non-human studies, non-English language articles, and unavailable full-text articles were excluded.

Results:

There were 2408 patients with neurogenic cough in this review, treated with medical therapy (77%), speech therapy (19%), both medical and speech therapy (1%), and procedural therapy (3%). The included studies ranged from low to intermediate quality. Overall, most interventions demonstrated successful improvement in cough. However, the heterogeneity of included study designs precluded direct comparisons between intervention types.

Conclusion:

This meta-analysis compared various treatments for neurogenic cough. Procedural therapy should be considered in the armamentarium of neurogenic cough treatments, particularly in patients refractory to, or intolerant of, the side effects of medical therapy. Lastly, this review illuminates key areas for improving neurogenic cough diagnosis, such as strict adherence to diagnostic and treatment guidelines, sophisticated reflux testing, and standardized, consistent outcome reporting.

INTRODUCTION

Neurogenic cough (NC) affects approximately 11% of Americans¹ and negatively impacts quality of life, causing dysphonia, urinary incontinence, social anxiety, and depression.² NC theoretically arises from neural injury, with subsequent upper airway sensory nerve hypersensitivity. The reflexive coughing episodes characteristic of NC are typically triggered by innocuous stimuli: laughing, talking, changing body position, externally manipulating the neck, inhaling cold air,³ or in the setting of underlying reflux of non-acid or acid gastric contents.⁴ Despite its high prevalence and disruptive impact, current understanding of NC pathophysiology, accurate diagnosis, and appropriate treatments is relatively nascent.

In 2016, the American College of Chest Physicians (CHEST) Cough Expert Panel published evidence-based guidelines for diagnosing and treating “unexplained chronic cough,” defined as cough persisting longer than 8 weeks of unknown cause or refractory to standard therapy.⁵ The panel supported gabapentin therapeutic trials with appropriate risk–benefit assessment (Grade 2C, weak recommendation, low or very low-quality evidence⁶) and multimodality speech pathology therapy (Grade 2C). They recommended against treatment with a proton pump inhibitor (PPI) given negative workup for acid reflux disease (Grade 2C) or inhaled corticosteroids without evidence of bronchial hyperresponsiveness or sputum eosinophilia (Grade 2B, weak recommendation, moderate-quality evidence⁶).⁵ These guidelines established gabapentin as the gold-standard therapy, despite its potentially undesirable side effects: fatigue, dizziness, and dry mouth.⁷ Nevertheless, prescribing practices among otolaryngologists are still variable.⁸

While past studies have systemically reviewed both pharmacologic⁹ and non-pharmacologic (i.e. speech therapy [ST])¹⁰ interventions for NC, superior laryngeal nerve (SLN) block has recently emerged as a novel treatment. The internal branch of the SLN conveys sensory innervation from the laryngeal vocal folds and above.¹¹ SLN block – either via injection of local anesthetic with corticosteroid or surgical transection – could provide a potential therapeutic option for NC that avoids the side effects of commonly used medications. Other recently described procedural treatments for NC include botulinum toxin (BTX) injection of laryngeal muscles¹² and vocal fold augmentation with temporary filler, such as methylcellulose or hyaluronic acid.¹³

Given these additional NC treatments, an updated review of interventions is necessary. We aim to describe the clinical effectiveness of pharmacologic and nonpharmacologic interventions for NC, particularly how SLN block compares to established pharmacologic treatments.

METHODS

PRISMA guidelines were followed and a review protocol was published on PROSPERO (ID: CRD42020171753). This review was exempt from the Washington University Institutional Review Board. Studies were selected using the PICOS (population, interventions, comparators, outcomes, and study design) format: 1) population: adults (≥18 years old) with NC; 2) intervention: any pharmacologic or nonpharmacologic treatment for NC; 3) comparator: any relevant interventions and/or single cohorts; 4) outcome: cough-specific quality of life (QoL) outcomes; 5) study design: clinical trials, cohort studies, case-control studies, and case series with ≥10 subjects. Active smokers, case reports, review articles, non-human participant studies, non-English language articles, and unavailable full-text articles were excluded (Figure 1). For a study to meet inclusion criteria, NC was operationally defined along CHEST guidelines: cough persisting ≥8 weeks after ruling out other etiologies, or after failure of supervised therapeutic trial(s) for the most common causes of persistent cough (pulmonary disease, sinonasal disease, and reflux-associated disease).⁵

Fig. 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram of studies through systematic review.

Literature Search

A medical librarian (A.H.) searched the literature for the concepts of chronic cough and medical, surgical, and other treatment modalities and related synonyms. The search strategies were executed in Ovid-Medline 1946–2020, Embase.com 1947–2020, Scopus 1823–2020, Cochrane Library, and Clinicaltrials.gov. All searches were completed by February 2020, yielding 2122 unique citations for analysis. Reference lists were screened for any relevant articles not found in the initial literature search. Complete search strategies are included in Supporting Information 1.

Study Selection

Four authors (a.m.p., j.j.l., l.j., n.s.w.) independently screened each study twice for eligibility. Titles and abstracts were first reviewed for inclusion, followed by full texts. Any discrepancies (21 total) were addressed and resolved by consensus with the senior author (m.n.h.).

Data Extraction and Summary Measures

Reviewers (a.m.p., j.j.l., l.j., n.s.w.) worked independently to extract the data, which was cross-checked twice (n.s.w.). If data were not available in the published article, corresponding authors were contacted via e-mail. The primary outcome was cough-specific QoL, which included validated patient-reported outcome measures (PROMs). The secondary outcome was any adverse event (AE).

Statistical Plan

Descriptive statistics characterized the study population, intervention type, cough-specific QoL outcomes used, and frequency of AEs. CHEST recommends the use of patient-reported QoL to assess the efficacy of NC treatments.⁵ Thus, meta-analysis was performed for studies implementing similar cough-specific PROMs, analyzing pre-/post-treatment mean difference or post-treatment mean scores. The relative risk of AEs was calculated. Data were analyzed as intention-to-treat (ITT) and stratified by type of intervention: medical, speech, or procedural therapy. Sensitivity analyses were performed to assess the impact of removing studies with missing data or high risk of bias on the results. Most crossover randomized controlled trials (RCTs) only reported results at the first time point, rather than at the end of the trial.¹⁴ Heterogeneity was calculated using I².¹⁴ Given the inclusion of observational studies, we anticipated wide study heterogeneity and planned to use a random effects model.¹⁵ Funnel plots assessed for publication bias.¹⁶ Statistical analysis was performed with STATA, version 16.1 (StatCorp, LLC, College Station, TX).

Quality Assessment

Reviewers (a.m.p., j.j.l., l.j., n.s.w.) independently assessed the quality of observational studies using the Methodological Index for Non-Randomized Studies (MINORS) criteria¹⁷ and assessed the risk of bias of RCTs with the revised Cochrane Risk of Bias 2 (RoB 2).¹⁸ The MINORS criteria assigns a maximum sum score of 16 for non-comparative and 24 for comparative studies; higher scores indicate higher study quality.¹⁷ The RoB 2 assigns categories of “low,” “some,” and “high” risk of bias.

RESULTS

Forty-four articles^{2,7,11,12,19–58} were included after initial screening; seven manuscripts^13,59–64 were added after reviewing their reference lists, totaling 51 manuscripts for systematic review. Thirty-seven articles^{2,7,11–13,19,21–23,25,29–33,35–44,46–51,54,55,57,58,60,63} had sufficient information to perform meta-analysis (Figure 1). There were 30 RCTs^{7,19,20,22,23,25–28,30,32,34–37,39,40,42–45,48,50,51,53,56,57,59,62,64} (including three unpublished completed clinical trials^20,27,34), 10 retrospective case series,^{2,11–13,31,46,49,55,58,63} nine prospective case series,^{21,29,33,38,47,52,54,60,61} one prospective cohort study,²⁴ and one retrospective cohort study.⁴¹ For studies that produced more than one manuscript, each article was assessed for unique patient data.^28,40,53,64

Quality of Included Studies

Twenty-one observational studies^{2,11–13,21,24,29,31,33,38,41,46,47,49,52,54,55,58,60,61,63} were assessed with the MINORS criteria (Supporting Information 2). Eighteen non-comparative studies^{2,11–13,21,29,31,33,38,46,47,49,54,55,58,60,61,63} had sum scores in the middle-to-higher range (7–14) and three comparative studies^24,41,52 had sum scores in the middle-range (14–16). Most of the RCTs^{7,19,20,22,25–28,30,32,34–37,39,40,42–45,48,50,51,53,56,57,59,62,64} assessed with the RoB 2 ranged from some to high concern for bias; only one RCT²³ had low concern for bias (Supporting Information 3).

Systematic Review

Despite strict inclusion criteria, there was large inconsistency in the diagnosis of NC across studies. For example, 24 unique terms labeled the condition; the most common was “chronic cough” (10 studies). Two studies used the term “neurogenic cough,”^12,29 one study used “neurogenic chronic cough,”⁵⁸ and one study used “chronic neurogenic cough.”¹¹ Furthermore, only three studies reported using CHEST guidelines to diagnose patients.^12,24,48

There were 2408 patients enrolled; 2196 (91%) completed the trials. Of the 2341 patients with known sex, the majority were women (1618 patients, 69%). For studies that reported either the mean or median age of patients, most were 60–69 years old (32 studies) (Table I).

TABLE I.

Characteristics of Included Studies, Stratified by Intervention Type: Medical Therapy, Speech Therapy, Procedural Therapy.

Author or clinical trial number (year)	Location	Study design	# Patients enrolled; completed therapy (males/females)	Average age in years	Experimental intervention (n patients)	Comparator intervention (n patients)	Treatment duration	Primary outcome measure	Secondary outcome measure(s)
Medical therapy
Pizzichini et al (1999)	Hamilton, Ontario, CAN	Parallel RCT	48; 44 (16/28)	E: 43 C: 47	Inhaled corticosteroid: budesonide 400 μg/inhalation BID (24)	Placebo inhaler, 1 inhalation BID (24)	2 weeks	Cough discomfort VAS	Sputum examination
Chaudhuri et al (2004)	Glasgow, Scotland, UK	Crossover RCT	93; 88 (32/57)	59	Inhaled corticosteroid: fluticasone 500 μg/inhalation BID	Placebo inhaler, 1 inhalation BID	2 weeks	Cough severity VAS	Sputum examination Serum examination Exhaled NO and CO
Jeyakumar et al (2006)	Cleveland, OH, USA	Parallel RCT	28; 28 (13/15)	E: 54.6 C: 49.7	Tricyclic antidepressant: amitriptyline 10 mg qHS (15)	Codeine/guaifenesin 100 mg/5 mL, 10 mL q6hr (13)	10 days	Reduction in cough frequency and severity	CQLQ
Morice et al (2007)	East Yorkshire, England, UK	Crossover RCT post hoc analysis of AEs	27; 27(9/18)	55	Opioid analgesic: morphine 5 mg BID	Placebo oral tablet BID	4 weeks	LCQ Number of patients with sedation	PFTs Citric acid cough challenge
Dickinson et al (2014)
Ribeiro et al (2007)	Sao Paulo, Brazil	Parallel RCT	64; 64 (22/42)	E: 46 C: 50	Inhaled corticosteroid: beclomethasone 500 μg/inhalation TID (44)	Placebo inhaler, 2 inhalation TID (20)	2 weeks	Cough symptom diary	Cough severity VAS
Yousaf et al (2009)	Leicester, England, UK	Parallel RCT	30; 28 (6/24)	E: 63 C: 61	Macrolide antibiotic: erythromycin 250 mg daily (15)	Placebo oral tablet daily (15)	12 weeks	24-hr cough frequency (coughs/hr)	LCQ Cough VAS Capsaicin cough challenge Sputum neutrophil count
Qiu et al (2010)	Shanghai, China	Parallel RCT	240; 214 (93/147)	E: 45 C: 45	Bronchodilator and antihistamine: diprophylline 200 mg TID + cetirizine 10 mg daily Corticosteroid: oral prednisone 25 mg daily ×1 week + inhaled budesonide 200 μg/inhalation BID PPI and antimotility agent: omeprazole 20 mg BID + domperidone 10 mg TID (120)	Bronchodilator and antihistamine Methoxyphenamine 2 capsules TID + cetirizine 10 mg daily Corticosteroid: oral prednisone 25 mg daily ×1 week + inhaled budesonide 200 μg/inhalation BID PPI and antimotility agent omeprazole 20 mg BID + domperidone 10 mg TID (120)	8 weeks	Cough symptom score	Rate of cough control LCQ
Ryan et al (2012)	Newcastle, NSW, Australia	Parallel RCT	62; 52 (22/40)	E: 62.7 C: 60.9	Neuromodulator: Gabapentin titrated up to 1800 mg daily, then tapered off (32)	Placebo oral tablets (30)	12 weeks	LCQ	Cough frequency (cough/hr) Cough symptom score Urge-to-cough score Laryngeal dysfunction score
Khalid et al (2014)	Manchester, England, UK	Crossover RCT	21; 19 (6/15)	53	TRPV-1 inhibitor: SB-705498	Placebo oral tablet	Single dose	Capsaicin cough challenge	24-hr cough frequency (coughs/hr) Cough severity VAS Urge-to-cough VAS CQLQ
Ternesten-Hasseus et al (2014)	Gothenburg, Sweden	Crossover RCT	24; 22 (2/22)	52	Capsaicin capsule 0.4 mg BID × 2 weeks, then 0.8 mg BID × 2 weeks	Placebo oral tablet daily	4 weeks	Capsaicin cough challenge	HRCQ Cough symptom diary
Abdulqawi et al (2015)	Manchester, England, UK	Crossover RCT	24; 18 (6/18)	54.5	P2X3 inhibitor: AF-219 600 mg BID	Placebo oral tablet BID	2 weeks	Daytime cough frequency (coughs/hr)	24-hr and nighttime cough frequency (coughs/hr) CQLQ Cough severity VAS Urge-to-cough VAS Cough global rating of change
NCT02397460 (2015)	UK	Crossover RCT	24; 24 (sex NR)	NR	P2X3 inhibitor: AF-219 50 mg (12 patients), AF-219 300 mg (12 patients)	Placebo oral tablet	Single dose	Daytime cough frequency (coughs/hr)	Capsaicin and ATP cough challenge Urge-to-cough VAS
Hodgson et al (2016)	Nottingham, England, UK	Parallel RCT	44; 20 (14/30)	E: 59.6 C: 56.9	Macrolide antibiotic: azithromycin 500 mg daily for 3 days, then 250 mg TID for remainder of treatment period (21)	Placebo oral tablet daily for 3 days, then TID for remainder of treatment period (21)	8 weeks	LCQ	Cough severity score FeNO
Belvisi et al (2017)	London, England, UK	Crossover RCT	20; 18 (5/15)	63.1	TRPV-1 inhibitor: XEN-D0501 4 mg BID	Placebo oral tablet BID	2 weeks	Daytime cough frequency (coughs/hr)	Cough severity VAS Urge-to-cough VAS LCQ Capsaicin cough challenge Cough global rating of change
Jang et al (2017)	Boston, MA, USA	Parallel RCT	30; 19 (11/7)	E: 42 C: 49	Tricyclic antidepressant: amitriptyline 12.5 mg qHS, maximum 50 mg daily (15)	Placebo oral tablet daily (15)	8 weeks	Modified RSI	VHI-10 Overall symptom severity Perceived degree (%) of symptom change
Birring et al (2017)	London, England, UK	Crossover RCT	28; 27 (6/21)	62	Mast cell inhibitor: sodium cromoglicate (PA101) 40 mg/inhalation TID	Placebo inhaler TID	2 weeks	Daytime cough frequency (coughs/hr)	LCQ Cough severity VAS PFTs FeNO
Sadeghi et al (2018)	Cottingham, England, UK	Parallel RCT	50; 47 (17/32)	62	Leukotriene antagonist: montelukast 10 mg qHS (35)	Oral corticosteroid prednisolone 20 mg daily (15)	2 weeks	24-hr cough frequency (coughs/hr)	HARQ LCQ Sputum examination PFTs FeNO
NCT03282591 (2018)	UK, USA	Parallel RCT	185; 176 (41/135)	E: 62.7 C: 62.6	NK-1 inhibitor: serlopitant 5 mg daily (92)	Placebo oral tablet daily (93)	12 weeks	Daytime cough frequency (coughs/hr)	24-hr cough frequency (coughs/hr) Cough severity VAS LCQ
Dong et al (2019)	Shanghai, China	Parallel RCT	234; 217 (89/145)	E: 45.2 C: 47.5	Neuromodulator: baclofen 60 mg daily (117)	Neuromodulator: gabapentin 900 mg daily (117)	12 weeks	Successful rate of cough resolution	Cough symptom score Capsaicin cough challenge GerdQ
Morice et al (2019)	Cottingham, England, UK	Crossover RCT	24; 24 (3/21)	61	P2X3 inhibitor: gefapixant 100 mg	Placebo oral tablet	Single dose	Capsaicin, ATP, citric acid, distilled water cough challenges	Cough severity VAS Urge-to-cough VAS HARQ Cough frequency (coughs/hr)
NCT03482713 (2019)	Japan	Parallel RCT	23; 23 (6/17)	E: 54.5 C: 57.2	P2X3 inhibitor: gefapixant 45 mg BID (11)	Placebo oral tablet BID (12)	2 weeks	24-hour cough frequency (coughs/hr)	Daytime cough frequency (coughs/hr)
Smith et al (2020)	Manchester, England, UK	Parallel RCT	253; 222 (60/193)	60.2	P2X3 inhibitor: gefapixant 7.5 mg BID (64) 20 mg BID (63) 50 mg BID (63)	Placebo oral tablet BID (63)	12 weeks	Daytime cough frequency (coughs/hr)	24-hr and nighttime cough frequency (coughs/hr) LCQ Cough severity VAS Cough global rating of change
Smith et al (2020)	Manchester, England, UK	Crossover RCT	Study 1: 29; 27 Study 2: 30; 29 (10/49)	Study 1: 63.2 Study 2: 60.2	P2X3 inhibitor: gefapixant Study 1: 50 mg/100 mg/150 mg/200 mg BID, dose escalated every 4 days Study 2: 7.5 mg/15 mg/30 mg/50 mg BID, dose escalated every 4 days	Placebo oral tablet BID	16 days	Daytime cough frequency (coughs/hr)	Cough severity VAS Daily cough symptom diary LCQ
Bowen et al (2018)	Cleveland, OH, USA	Prospective cohort	28; 28(12/16)	61	Tricyclic antidepressant: amitriptyline or nortriptyline maximum 50–60 mg daily (9)	Neuromoduiator gabapentin maximum 1800 mg daily (19)	6 months	LCQ
Lee et al (2005)	New York, NY USA	Prospective case series	20; 20 (8/12)	53	Neuromodulator: gabapentin maximum 900 mg daily or carbamazepine 100 mg TID (if gabapentin intolerable)	None	4 weeks	Subjective improvement (yes or no)	Laryngeal EMG Videostroboscopy
Bastian et al (2006)	Downers Grove, IL, USA	Prospective case series	12; 12 (8/4)	52	Tricyclic antidepressant: amitriptyline 10 mg qHS	None	3 weeks	Subjective cough reduction	Time (days) to maximal effect
Xu et al (2013)	Shanghai, China	Prospective case series	16; 12 (9/7)	47.8	Neuromodulator + PPI: baclofen 20 mg TID + omeprazole 20 mg BID	None	8 weeks	Cough symptom score	GerdQ Capsaicin cough challenge
Dion et al (2017)	New York, NY USA	Prospective case series	16; 16 (6/10)	63	Opioid analgesic: tramadol 50 mg TID PRN	None	2 weeks	CSI	LCQ
Hong et al (2019)	Anyang, Korea	Prospective case series	33; 33 (10/23)	47	Inhaled corticosteroid: fluticasone 250 μg/inhalation BID or budesonide 400 ng/inhalation BID	None	2 weeks	Degree of persistent cough compared to baseline
Smith et al (2020)	Manchester, England, UK	Prospective case series	13; 13 (2/11)	60.1	NK-1 inhibitor: orvepitant 30 mg daily	None	4 weeks	Daytime cough frequency (coughs/hr)	Nighttime cough frequency (coughs/hr) CQLQ Cough severity VAS Cough global rating of change
Norris et al (2010)	Jackson, MS, USA	Retrospective cohort	11; 11 (1/10)	E: 67.3 C: 54.3	Tricyclic antidepressant: amitriptyline 25 mg qHS, maximum up to 50 mg daily neuromodulator: gabapentin or pregabalin, dose NR (if amitriptyline intolerable) (8)	Reflux regimen: PPI/antihistamine and lifestyle modifications (3)	1 or more months	Subjective improvement (yes or no)
Halum et al (2009)	Indianapolis, IN, USA	Retrospective case series	12; 10 (6/6)	58	Neuromodulator: pregabalin started at 75 mg BID, maximum 150 mg BID	None	4 weeks	Subjective rating of chief complaint	RSI GFI
Van de Kerkhove et al (2012)	Leuven, Belgium	Retrospective case series	51; 35 (10/41)	47	Neuromodulator: gabapentin 300 mg daily, maximum 600 mg BID	None	4 weeks	Cough severity score	Modified LCQ Number of patients improving on cough severity score
Stein et al (2013)	Boston, MA, USA	Retrospective case series	66; 66 (18/48)	NR	Tricyclic antidepressant: amitriptyline 10 mg daily (minimum)	None	3 months	Subjective response to therapy
Bastian et al (2015)	Downers Grove, IL, USA	Retrospective case series	32; 32 (8/24)	63	1st line tricyclic antidepressant: amitriptyline 10 mg daily, maximum 80 mg daily (<60 yo); desipramine 10 mg daily, maximum 80 mg daily (>60 yo) 2nd line neuromodulator: gabapentin 300 mg daily, maximum 2400 mg daily 3rd line selective serotonin reuptake inhibitor: citalopram 4th line neuromodulator: pregabalin 5th line neuromodulator: oxcarbazepine if failed all the above: capsaicin spray	None	6 or more months	Percent reduction of cough symptoms from baseline
Zalvan et al (2019)	Valhalla, NY, USA	Retrospective case series	29; 29 (9/20)	63	Trigger reduction method: plant-based diet, standard reflux precautions, nasal saline irrigation 4–5 times daily, intranasal steroid/antihistamine combination	None	6 weeks	RSI CSI	Percent reduction in RSI, CSI from baseline Number of patients who are clinical responders (reduction in RSI by ≥6 points)
Vertigan et al (2016)	Newcastle, NSW, Australia	Parallel RCT	40; 35 (13/27)	E: 61 C: 64	Neuromodulator: pregabalin titrated up to 300 mg daily, then tapered off with speech pathology treatment (5 sessions) (20)	Placebo oral tablet with speech pathology treatment (5 sessions) (20)	14 weeks	LCQ	Cough severity VAS 24-hr cough frequency (coughs/hr) Capsaicin cough challenge Urge-to-cough scale CAPE-V, DSI, LHQ, VHI
Speech therapy
Vertigan et al (2006)	Newcastle, NSW, Australia	Parallel RCT	97; 87 (23/64)	59.4	SPEech Pathology Intervention Program for CHronic Cough (SPEICH-C) 30-min sessions (43)	Healthy Lifestyle Education intervention program (HLE) 30-min sessions (44)	4 sessions	Symptom severity and frequency rating	Limitation of symptoms on everyday activity Clinician assessment of participant ability to understand and implement strategies
Vertigan et al (2008)	Newcastle, NSW, Australia	Parallel RCT	97; 83 (22/61)	E: 58.9 C: 61.5	SPEICH-C 30-min sessions	HLE 30-min sessions	4 sessions	Perceptual voice analysis rated by SLPs	Acoustic and electroglottographic analyses
Young et al (2009)	Manchester, England, UK	Parallel RCT	30; 30 (10/20)	E: 60.2 NC: 54.2 PC: 61.1	Mindfulness based-stress reduction 30 min daily (10)	NC: no intervention (11) PC: voluntary suppression (9)	7–10 days	Citric acid cough challenge	Urge-to-cough VAS
Chamberlain et al (2017)	London, England, UK	Parallel RCT	75; 63 (24/51)	E: 61 C: 56	Physiotherapy, speech and language therapy intervention (PSALTI) 45-min sessions (34)	One-on-one standardized healthy lifestyle advice 45-min sessions (41)	4 sessions	LCQ	24-hr cough frequency (coughs/hr) Cough severity VAS VPQ
Kapela et al (2019)	Newcastle, NSW, Australia	Parallel RCT	18; 15 (2/16)	E: 59 C: 57	Standard speech pathology treatment and supplemental prerecorded videos for home practice 30–45-min sessions (9)	Standard speech pathology treatment 30–45-min sessions (9)	2 or more sessions	Accuracy of therapy technique rated by SLP	Symptom frequency and severity rating scale LCQ CAPE-V
Ryan et al (2010)	Newcastle, NSW, Australia	Prospective case series	17; 17 (8/9)	61	Speech pathology program for chronic cough 30-min sessions	None	Up to 4 sessions	Capsaicin cough challenge	Urge-to-cough VAS Cough frequency in 1-hour LCQ
Patel et al (2011)	London, England, UK	Prospective case series	23; 23 (10/13)	60	Cough suppression physiotherapy program	None	Up to 3 sessions	LCQ	Subjective reporting of cough frequency and sleep disturbance
Vertigan et al (2017)	Newcastle, NSW, Australia	Prospective case series	33; 33 (11/22)	62.6	SPEICH-C	None	5 sessions	Vocal loading test completion, accuracy, effect on cough, effect on acoustic measures
Yang et al (2019)	Loma Linda, CA, USA	Retrospective case series	27; 27 (6/21)	62.4	Breath training therapy 45-min sessions	None	2–4 sessions	CSI	Maximum phonation time Aerodynamic measures
Procedural therapy
Sasiesta et al (2016)	Rochester, MN, USA	Retrospective case series	23; 21 (2/19)	61	EMG-guided laryngeal botulinum toxin A injection	None	1 or more injections	Subjective cough severity	Number of injections
Litts et al (2018)	Aurora, CO, USA	Retrospective case series	23; 23 (12/11)	67.7	Vocal fold injection augmentation with methylcellulose or hyaluronic acid	None	1 injection	CSI	Number of patients reporting improvement of symptoms
Simpson et al (2018)	Dallas, TX, USA	Retrospective case series	18; 18 (3/15)	60	Local anesthetic + corticosteroid injection: 1% lidocaine with epinephrine (1:100,000) or 0.5% bupivacaine + triamcinolone acetonide 200 mg/5 mL or methylprednisolone 80 mg/1 mL	None	1 or more injections	CSI	Number of injections
Dhillon (2019)	Bethesda, MD, USA	Retrospective case series	10; 10 (3/7)		Local anesthetic + corticosteroid injection: 1% lidocaine with epinephrine (1:100,000) + triamcinolone acetonide 200 mg/5 mL	None	1 or more injections	CSI	Number of injections Subjective reporting of improvement

AE = adverse event; BID = twice daily; C = comparator intervention; CAPE-V = Consensus Auditory-Perceptual Evaluation of Voice; CO = carbon monoxide; CQLQ = Cough-specific Quality of Life Questionnaire; CSI = Cough Severity Index; DSI = Dysphonia Severity Index; E = experimental intervention; EMG; electromyography/electromyographic; FeNO = fraction of exhaled nitric oxide; GFI = Glottal Function Index; HARQ = Hull Airways Reflux Questionnaire; HRCQ = Hull Reflux Cough Questionnaire; LCQ = Leicester Cough Questionnaire; LDQ = Laryngeal Dysfunction Questionnaire; LHQ = Laryngeal Hypersensitivity Questionnaire; MINORS = Methodological Index for Non-Randomized Studies; NC = negative comparator; NK-1 = neurokinin-1; NO = nitric oxide; NR = not reported; PC = positive comparator; PFTs = pulmonary function tests; PRN = pro re nata (as needed); q6hr = every 6 hours; qHS = every bedtime; RCT = randomized controlled trial; ROB 2 = revised Cochrane Risk of Bias 2; RSI = Reflux Symptom Index; SLP = speech-language pathologist; TID = three times daily; TRPV-1 = transient receptor potential vanilloid-1; VAS = visual analog scale; VHI = Voice Handicap Index; VPQ = Vocal Performance Questionnaire; yo = years old.

Three categories of intervention were assigned: medical therapy (1857 patients, 77%, 38 studies), ST (457 patients, 19%, 9 studies), and procedural therapy (73 patients, 3%, 4 studies). One study⁵¹ investigated both medical therapy and ST (20 patients, 1%) (Table I).

Various medical therapies were prescribed: neuromodulating drugs (NMDs, i.e., gabapentin, pregabalin, baclofen), tricyclic antidepressants (TCAs), inhaled corticosteroids, opioids, macrolide antibiotics, PPIs, and investigational drugs (i.e., transient receptor potential vanilloid-1 [TRPV-1] inhibitors). Two studies included medical therapy that may be considered “alternative” or “homeopathic” medicine: oral capsaicin⁵⁰ and “trigger reduction method” (plant-based diet, reflux precautions, nasal saline irrigation, and intranasal corticosteroid or antihistamine).⁵⁸ The most commonly studied medical therapies were investigational drugs (11 studies) and gabapentin (7 studies). The median and mode duration of medical therapy was 4 and 2 weeks, respectively.

Next, ST commonly involved education about the harmful impact of cough, cough suppression techniques (breathing exercises, mindfulness training, voice therapy, etc.), and stress or anxiety counseling. The median treatment duration for ST was four sessions.

Finally, procedural therapies included SLN block via injection of local anesthetic (lidocaine or bupivacaine) and corticosteroid (triamcinolone acetonide or methylprednisolone),^11,46 bilateral thyroarytenoid BTX injection,¹² and vocal fold augmentation with methylcellulose or hyaluronic acid.¹³ Among the studies of procedural therapy, there was a median of one treatment. Specifically for SLN block, the mean number of treatments was 2.3¹¹ and 2.4⁴⁶ injections.

The three most commonly used cough-specific PROMs were the Leicester Cough Questionnaire (LCQ, 19 studies),⁶⁵ Cough Severity Index (CSI, 6 studies),⁶⁶ and Cough-specific Quality of Life Questionnaire (CQLQ, 4 studies).⁶⁷ Only 12 studies reported voice-related outcomes (i.e., Consensus Auditory-Perceptual Evaluation of Voice,⁶⁸ Voice Handicap Index⁶⁹); among these studies, 10 different outcome metrics were used (Table I).

Meta-Analysis

The primary outcome of cough-specific QoL was analyzed first for studies reporting the most commonly used PROMs: LCQ and CSI. Next, studies dichotomizing outcomes into cough “improvement” and “no improvement” were analyzed. The secondary outcome was the proportion of patients with at least one AE. Meta-analysis was conducted for active treatments versus placebo and for active treatment groups only.

Leicester Cough Questionnaire.

The LCQ is a validated, cough-specific PROM of 19 items rated on a 7-point Likert scale. The total score ranges from 3 to 21; higher scores indicate improved QoL.⁶⁵ The LCQ has a pre-/post-treatment minimal clinically important difference (MCID) score of 1.3.⁷⁰

Active treatments versus placebo.

Overall, medical therapy and ST were associated with a significant pre-/post-LCQ mean difference over placebo (1.60, 95% CI 0.78–2.42, I² = 0%, n = 4 studies). Although morphine had the largest LCQ mean difference over placebo among the medical therapies, this study had high risk of bias.⁴⁰ All of these studies had missing patient data, precluding sensitivity analysis (Supporting Information 4).

Active treatments only.

When excluding placebo comparisons, pre–post-LCQ mean difference was significant for gabapentin⁷ and “three-step empirical therapy” (a combination of a novel bronchodilator (diprophylline or methoxyphenamine) with an oral antihistamine [step 1], oral and inhaled corticosteroids [step 2], and a PPI with an antimotility agent [step 3]).⁴³ ST was also associated with an improved LCQ mean difference.^25,37 All of these studies had missing patient data and similar risk of bias, precluding sensitivity analysis (Figure 2).

Fig. 2.

Forest plot of the meta-analysis of pre-/post-Leicester Cough Questionnaire mean difference scores for active treatment groups, stratified by intervention type (medical therapy and speech therapy). Reference line indicates the minimal clinically important difference, 1.3. LCQ = Leicester Cough Questionnaire; PSALTI = physiotherapy, speech and language therapy intervention; ST = speech therapy.

Cough Severity Index.

The CSI is a validated, cough-specific PROM with 10 items rated on a 5-point Likert scale. The total score ranges from 0 to 40, with lower scores indicating improved QoL. While there is no established pre-/post-treatment MCID, a total score > 3.23 is considered “symptomatic for cough.”⁶⁶ No placebo-controlled RCTs utilized the CSI.

In studies of active treatments, Figure 3 compares pre-/post-CSI mean differences and Figure 4 compares post-intervention mean CSI scores. Compared to tramadol, SLN block^11,46 had a greater decrease in CSI score, indicating greater symptom relief. While all interventions led to decreased CSI scores, all of the post-treatment CSI scores were >3.23. Thus, although cough QoL improved, cough was not, on average, completely resolved and patients were still subjectively symptomatic. Of note, trigger reduction therapy⁵⁸ and breath training therapy⁵⁵ led to lower post-treatment CSI than the other therapies. However, pre-treatment CSI was also lower for these interventions compared to the other therapies, suggesting that these patients were not as severely symptomatic at baseline. None of these studies had missing patient data and all had similar MINORS scores (ranged 7–11 out of 16), precluding sensitivity analysis.

Fig. 3.

Forest plot of the meta-analysis of pre-/post-Cough Severity Index mean difference scores for active treatment groups, stratified by intervention type (medical therapy and procedural therapy). CSI = Cough Severity Index; SLN = superior laryngeal nerve.

Fig. 4.

Forest plot of the meta-analysis of Cough Severity Index post-treatment scores for active treatment groups, stratified by intervention type (medical therapy, procedural therapy, and speech therapy). Reference line indicates the minimal score considered for symptomatic cough, 3.23. CSI = Cough Severity Index; SLN = superior laryngeal nerve; VF = vocal fold.

Improvement in cough symptoms.

Definitions of “improvement” varied from cut-offs based on validated PROMs to subjective reporting of improved symptoms by patients (Supporting Information 5).

Active treatments versus placebo.

Patients receiving medical therapies were more than twice as likely to report improved cough compared to placebo (relative risk [RR] 2.17, 95% CI 1.02–4.60, I² = 57%, n = 3 studies). The study with inhaled beclomethasone had the largest effect, but was at high risk of bias.⁴⁴ Two of these studies^7,35 had missing patient data, and two studies^35,44 had similar RoB 2 scores, precluding sensitivity analysis (Supporting Information 6).

Active treatments only.

Medical therapy was associated with 60% (95% CI 52–68%, I² = 73%, n = 20 studies) of patients reporting improved cough; however, individual studies had widely variable effects. In Bastian et al’s²¹ retrospective case series, 41% of participants initially treated with TCAs (amitriptyline or desipramine) eventually switched to NMDs (gabapentin, pregabalin, or oxcarbazepine) because of persistent cough. Furthermore, gabapentin’s effect on cough improvement varied greatly; some studies showed a majority of patients had improved,^21,41 while in others, a low proportion of patients improved.^7,30,38,63 ST had the greatest proportion of patients with improved cough (86%, 95% CI 75–95%, n = 2 studies) (Figure 5).

Fig. 5.

Forest plot of the meta-analysis of the proportion of patients with cough improvement for active treatment groups only, stratified by intervention type. The medical therapy category is ordered in clusters by drug class: neuromodulating drugs, tricyclic antidepressants, opioid-cough suppressant, inhaled corticosteroids, and combination-drug regimens. Effect sizes (ES) closer to one indicate a higher proportion of patients with cough improvement and are considered to favor treatment. BTX = botulinum toxin; PPI = proton pump inhibitor; VF = vocal fold.

Two procedural therapy studies had vastly different effects on cough improvement. Only 43% of patients reported improved cough after one bilateral thyroarytenoid BTX injection, and 36% of patients requested additional BTX injections.¹² In contrast, vocal fold augmentation improved cough for 78% of patients, yet 35% requested thyroplasty afterwards.¹³ Sensitivity analysis excluding studies with missing data showed a similar proportion of patients with improved cough for all interventions overall (72%, 95% CI 64–81%, I² = 64%, n = 16 studies), and for medical therapy (69%, 95% CI 59–79%, n = 1 study).

Adverse Events

Active treatments versus placebo.

Overall, there was a higher risk of AEs in treatment groups getting medical therapy than placebo (RR 1.93, 95% CI 1.22–3.07, I² = 81%, n = 12 studies) (Supporting Information 7).

Much of the weight contributing to this effect is due to the studies of investigational drugs: oral P2X3 inhibitors (AF-219¹⁹ and gefapixant^39,48) and an oral TRPV-1 inhibitor (XEN-D0501).²² P2X3 inhibitors block receptors in airway primary sensory nerves. The most common AE experienced by P2X3 inhibitor groups was taste disturbance, presumably from co-expression of P2X3 receptors in taste afferent nerves.⁴⁸ In three RCTs, all or nearly all participants taking the P2X3 inhibitor experienced taste disturbance compared to zero patients assigned to placebo, and this was dose-dependent.^19,39,48 The predominant AE of taste disturbance was tempered by improvement in cough-specific PROMs over placebo for all three trials (insufficient data for meta-analysis).^19,39,48 XEN-D0501 inhibits TRPV-1 ion channel receptors on vagal afferent nerves. Its most common AEs were taste disturbance and body temperature disturbance, but unlike the P2X3 inhibitors, XEN-D0501 led to “no significant” improvements in cough-specific PROMs versus placebo.²²

There were no AEs in both treatment and placebo groups for ST.²⁵ Sensitivity analysis excluding studies with missing data or high risk of bias showed a similar overall AE rate (RR 1.15, 95%CI 0.58–2.26, I² = 36%, n = 4 studies), however, this analysis led to exclusion of the ST studies.

Active treatments only.

Of the NMDs, gabapentin had a lower AE rate than pregabalin or baclofen. Common AEs of NMDs were nausea,⁷ fatigue,^7,63 sedation or somnolence,^30,31,54 and dizziness.^7,30,51 The AE rates for amitriptyline were highly variable. Two studies^21,41 had low AE rates (dry mouth); however, one study³⁵ with a higher AE rate for amitriptyline had high risk of bias, and in another study,⁴¹ 25% of patients switched from amitriptyline to gabapentin because of persistent cough or intolerable side effects. The AE rate for alternative treatments varied from 0% for anti-reflux/anti-histamine therapy,⁴¹ to 55% for three-step empirical therapy (most commonly drowsiness)⁴³ to 88% for oral capsaicin (most commonly hoarseness).⁵⁰ Investigational drugs mostly had high AE rates. In addition to P2X3 and TRPV-1 inhibitors, orvepitant, (inhibits neurokinin-1 [NK-1] receptors, which hypothetically centrally modulate the cough reflex), had a 69% AE rate (most commonly fatigue, lethargy, and somnolence).⁴⁷ ST, whether individualized²⁵ or standardized,⁵¹ had zero AEs (Figure 6).

Fig. 6.

Forest plot of the meta-analysis of the proportion of patients with at least one adverse event (AE) for active treatment groups only, stratified by intervention type. The medical therapy category is ordered in clusters by drug class: neuromodulating drugs, tricyclic antidepressants, opioids, inhaled corticosteroids, macrolide antibiotic, combination-drug regimens, alternative medicine, and lastly investigational drugs. Effect sizes (ES) closer to 0 indicate a lower proportion of patients with at least 1 AE and is considered to favor treatment. BTX = botulinum toxin; PPI = proton pump inhibitor; PSALTI = physiotherapy, speech and language therapy intervention; SLN = superior laryngeal nerve; SPT = speech pathology treatment; VF = vocal fold.

Procedural therapy studies generally had low AE except for bilateral thyroarytenoid BTX injection: temporary liquid dysphagia (62%) and dysphonia (90%).¹² Zero AEs were reported for vocal fold augmentation.¹³ Patients who received SLN block experienced brief laryngospasm (1 out of 18 patients⁴⁶) and temporary throat paresthesia (1 out 18 patients⁴⁶ and 2 out of 10 patients¹¹). No serious AEs (i.e., death, aspiration pneumonia) were reported in any of the procedural therapy studies.

Sensitivity analysis excluding studies with missing data or high risk of bias showed lower AE rates for medical therapy (24% 95% CI 1–58%, I² = 93%, n = 8 studies), for procedural therapy (8%, 95% CI 0–30%, n = 3 studies), and for medical and procedural therapy combined (19%, 95% CI 3–44%, I² = 92%, n = 11 studies); however, this analysis led to exclusion of the ST studies.

Publication bias.

Studies reporting the proportion of patients with improved cough (outcome with the highest number of applicable studies) were assessed with a funnel plot (Supporting Information 8), which did not show asymmetry, suggesting a low risk of publication bias.

DISCUSSION

This review examined 2408 patients with NC in studies of low-to-intermediate quality. Most patients were female, aged 60–69 years old. The most commonly used intervention was medical therapy, specifically gabapentin and investigational drugs. Overall, most interventions improved cough; however, almost all studies lacked long-term follow-up.

Medical therapy, ST, and procedural therapy were compared by analyzing LCQ, CSI, and the proportion of patients reporting cough improvement. Medical therapy, specifically gabapentin, TCAs, and P2X3-inhibitors, led to an improvement in cough and cough-specific QoL. ST, administered as regimens of education about the counter productivity of NC, cough substitution tactics (i.e., sipping water, talking through the cough, or breathing exercises), and counseling for emotional duress, was consistently favorable in improving cough and cough-specific PROMs. Yet, it is unknown which specific element of ST and by what mechanism of action leads to improved cough.²⁵ The appropriate treatment duration and whether or not ST should be standardized or individualized is also unknown. For procedural therapy, SLN block showed improvement in cough-specific QoL compared to vocal fold augmentation and laryngeal BTX injection; however, patients requested or received repeat treatments for all three procedures.

Medical therapy interventions had highest rate of AEs. This effect was largely due to the P2X3 and TRPV-1 inhibitors. These investigational drugs caused noticeable changes in taste and temperature sensations, respectively, despite any improvements in subjective cough. Even unpublished RCTs of P2X3 inhibitors^20,27 reported a high incidence of taste disturbance. Although not serious, the duration and patient tolerance of taste disturbance are unclear, as the longest treatment interval was 12 weeks.⁴⁸ Deeper understanding of optimal dosing, treatment duration, and patient tolerability is needed before clinical use of these investigational drugs. ST had no reported AEs, bolstering its use as an adjunct to medical therapy by the CHEST panel.⁵ Among the procedural therapies, bilateral thyroarytenoid BTX injection was associated with the most harm and the least benefit. Vocal fold augmentation, associated with a low AE rate in this meta-analysis, presents an indirect way of suppressing cough by decreasing glottic insufficiency. Patients are less likely to use increased glottic pressure to produce a strong voice, thereby avoiding phonotrauma that would hypersensitize nerves around the arytenoid mucosa and induce cough.¹³ AEs for SLN block were low, non-serious, and temporary. On average, about two injections were needed, but follow-up was not standardized.^11,46 Without controlled trials of SLN block, we cannot readily advocate for its superiority or non-inferiority over commonly used medical treatments.

Future research in developing an optimal treatment plan for NC involves understanding its pathophysiology, clinical presentation, accurate diagnosis, and appropriate end-point assessment of therapy. Historically, NC was diagnosed and treated via an anatomic–diagnostic protocol (ADP): systematically eliminating common causes for cough (i.e., reflux, asthma) by targeting different body systems. However, a meta-analysis of medical therapy for NC reported an approximate 40% failure rate with ADP.⁹ One theory for the failure rate of ADP is that traditional treatments were directed at incorrect aspects of NC pathophysiology. For example, PPIs lower the acidity, but not the volume, of gastric refluxate; thus, non-acid reflux may still activate cough receptors in the proximal esophagus and hypopharynx, causing NC.⁴ Another theory suggested by recent literature is that there are multiple factors for NC along the nervous system, from the cerebral cortex to cough receptors in respiratory mucosa. Disruption of any part of this laryngeal nerve afferent-efferent pathway may cause hypersensitivity leading to overactive coughing and, sometimes, motor neuropathy (vocal fold hypomobility).³ The goals of NC treatment have shifted to address these new targets via central neuromodulation (NMDs, TCAs, opioids) or direct inhibition of sensory receptors (P2X3, TRPV-1, and NK-1 inhibitors) or nerves (SLN block).

Clinically, NC predominantly affects middle-aged and older females, consistent with our review, but the pathophysiology of sex in NC remains unclear.⁷¹ In one study, sex hormones differentially affected cough reflex sensitivity during menstruation. However, this study’s population was comprised of premenopausal females,⁷² not the older perimenopausal or menopausal females typical of NC patients.

Many studies did not use standardized diagnostic criteria for NC; only three studies specifically state using the CHEST guidelines. Other similar directives, such as the British Thoracic Society guidelines⁷³ (used by four studies^22,25,48,57) exist. However, beyond the research setting, patients generally receive empiric therapy at their clinician’s judgment (often a PPI, which violates CHEST guidelines⁵) and may not undergo full diagnostic workup.⁸

Finally, there is a lack of consensus in outcome reporting. CHEST recommends patient-reported QoL, especially the LCQ, to assess treatment efficacy.⁵ Our review revealed an academic and geographic divide. The LCQ was used by pulmonologists, internists, and speech therapists internationally; the CSI was used by otolaryngologists in the United States. Both are advantageous in that they are cough-specific, conveniently administered, and measured on a continuous-level scale (thus are sensitive to change). The CSI was designed specifically for upper airway symptoms, unlike the LCQ;⁶⁶ thus, it may be more appropriate for measuring NC outcomes because it does not contain confounding questions specific to lower airway disease. In addition, there is some evidence that NC negatively impairs voice outcomes,^52,53 but inconsistent reporting prevented comparisons using this outcome. The tremendous social and functional impact of NC warrants standardized, consistent diagnosis, and PROMs to tailor therapy.

Despite a systematic search, our study has a number of limitations. The largest challenge was the lack of direct comparisons between different cough interventions, prohibiting network meta-analysis. Only one study compared two modalities, medical therapy (pregabalin) and ST.⁵¹ Study quality were generally low-to-intermediate, and specifically, procedural therapies were only investigated in small, unblinded single-cohort case series. Although common causes of chronic cough, such as asthma or reflux disease, were meticulously excluded, any impact from these challenging and heterogenous disease states was perhaps not absolutely eliminated due to the heterogeneity of study details – as mentioned, 24 different terms were used to describe NC among the 51 studies. Another limitation was variability in the dosage, titration, and follow-up of treatments, particularly for medical therapy. Eight RCTs^{7,32,35,43,48,50,51,62} and six observational studies^{2,24,31,38,41,49} included dose titration or dose escalation schedules (Table I). The overall effect of dosing regimens was obfuscated by short follow-up periods in the RCTs and by lack of consistency or individualized participant variability in dosing schedules in the observational studies. Because of short treatment duration and follow-up, longitudinal efficacy, and the long-term impacts of AEs for all three kinds of therapies (medical, speech, and procedural) could not be assessed. Finally, missing data from participants, entire cohort groups, or potentially eligible but non-English studies possibly impacted the conclusions of our review. Requests for missing data were made to all corresponding authors via e-mail. In all cases, the requested information was either unavailable, inadequate, or there was no reply.

The results of this review introduce SLN block as a potential treatment for NC. It is minimally invasive and avoids medical side effects; however, multiple injections may be required to achieve optimal symptom relief. The success of SLN block portends the potential of a more permanent option: SLN transection. In our own institution, we performed internal SLN transection on six patients with NC. All reported improvement on CSI; there were two complications – a self-resolved hematoma and dysphagia in a patient who had a previous Nissen fundoplication.⁷⁴ The efficacy of SLN block or transection should be further studied in a clinical trial of SLN block with placebo, compared to the gold standard, gabapentin, with an ethically responsible substitution for SLN block, measuring outcomes with the CSI.

CONCLUSION

Our meta-analysis described various treatments for NC, introduced future potential for procedural therapy, and highlighted areas for improving diagnosis of NC, such as strict adherence to CHEST guidelines or sophisticated non-acid reflux testing to rule out reflux-associated disease. NC is a frustrating condition for patients and physicians, who are often shuffling different medication trials with unclear certainty of improvement. Standardized reporting of outcomes for NC is critically important to inform clinicians, patients, and researchers managing this challenging clinical scenario.