INTRODUCTION

Cough is one of the most common reasons patients present to primary and secondary health care.1 Although often self-limiting, chronic cough causes substantial health burden and impaired quality of life.2-4 In the most recent European Respiratory Society and the American College of Chest Physicians guidelines, chronic cough in adults is defined as cough for more than 8 weeks duration, whereas earlier guidelines defined it as cough duration of more than 3 months.5, 6 Chronic cough is a significant public health burden, with prevalence ranging from 9% to 33% globally.7 This burden is even greater when considering that chronic cough can be the first symptom of chronic obstructive pulmonary disease (COPD),8 a major source of morbidity in adults.9, 10 Furthermore, chronic cough with phlegm is associated with increased all-cause mortality, even when lung function is not impaired.11-13

Although chronic cough is a common symptom of various diseases, such as asthma, rhinitis, post-nasal drip and gastro-oesophageal reflux disease (GORD),6 46% of patients with chronic cough presenting to specialist clinics do not have any apparent underlying condition even after clinical investigations.14 These patients are considered to have ‘undiagnosed’ or ‘idiopathic’ chronic cough.1 The diagnosis and management of patients with chronic cough are often challenging, largely due to insufficient understanding of the aetiology and pathological mechanisms, and lack of effective therapies. The concept of ‘cough hypersensitivity syndrome’ has also been proposed to describe hypersensitive cough responses, with the suggestion that this may arise from a neuropathic disorder.15

Many studies have recruited patients from clinics or certain industries, and evaluated risk factors for cough cross-sectionally. Findings of such studies are subject to multiple biases.16-20 Few population-based studies have investigated risk factors for chronic cough longitudinally12, 21-24 and their findings have not yet been systematically synthesized. Such a synthesis would extend understanding of the pathophysiological mechanisms, potential treatments and avenues for prevention for chronic cough. We conducted a systematic review to identify and synthesize the current evidence for longitudinal associations between a range of risk factors and chronic cough (with or without phlegm) in the general adult population.

METHODS

We followed the meta-analysis of observational studies in epidemiology (MOOSE) reporting guidelines for this review (Table S1 in the Supporting Information).25 Methods were specified and documented in a protocol a priori and published on PROSPERO (ID: CRD42020161973).

We systematically searched the MEDLINE and EMBASE electronic databases from inception to our last search date (2 May 2021). The search strategy included key words for risk factors and chronic cough (Table S2 in the Supporting Information).

Articles with longitudinal designs assessing risk factors for chronic cough, with and without chronic phlegm, were included. See Table S3 in the Supporting Information for detailed inclusion and exclusion criteria. For articles reporting chronic cough both with and without phlegm, both sets of results were extracted and treated separately. We included articles regardless of the measure of association used (i.e., OR, risk ratios or relative risks [RR] and hazard ratios [HR]).

Two authors (JZ and NSI) independently screened titles and abstracts of all identified articles after duplicates were removed. Eligible articles were retrieved in full text for further assessment. We screened the reference lists of these full texts for additional articles which met our criteria and consulted experts in the field for any grey literatures. Any disagreements during the process were settled by consultation and consensus with other authors (SCD, JLP or CJL).

Data extraction and statistical analysis

Relevant data were extracted using a pre-defined table (Table S4 in the Supporting Information). The quality of the articles was assessed using a form (Table S5 in the Supporting Information) based on the Newcastle–Ottawa scale (NOS).26

Articles were grouped according to the type of exposures. Forest plots were used to present results within similar exposure groups, but only estimates from ≥3 articles using the same effect estimate measure (i.e., OR, RR, HR or prevalence) and similar exposure measurements were meta-analysed. ORs and RRs underwent meta-analysis together following conversion: RR = OR ÷ (1 − p + (p × OR)), where p is the prevalence of chronic cough in the reference group. Adjusted effect estimates with 95% CIs were reported where available. If multiple results were reported from one article, only estimates with the most similar characteristics (i.e., set of confounders adjusted, age group and cough definitions) were pooled while other results were presented in tables or figures without pooling. Heterogeneity of associations was assessed using I2 and was considered minimal if I2 < 25%, moderate if 25% ≤ I2 ≤ 75% and substantial if I2 > 75%. As only a small number of articles were included , fixed-effect models were used,27, 28 funnel plots were not undertaken, heterogeneity and publication bias were explored narratively and interactions were listed in tables without further exploration. All statistical analyses were performed using STATA, version 16 (Stata Corp LP, College Station, TX, USA).

RESULTS Study selection and characteristics

Our search identified 4051 records from electronic databases (last performed on 2 May 2021) and an additional 105 records sourced from manual reference searches; no grey literature has been identified. No attempts were made to contact the authors of the records as all relevant information was available. Of the total articles identified, 26 met the eligibility criteria (Figure S1 in the Supporting Information).

The 26 eligible articles reported results from 19 studies, including two retrospective studies29, 30 and 17 prospective studies.31-51 Seven of these 19 studies generated multiple articles, all reporting different risk factors from the same cohort and there was no duplication in our analyses (Table 1).

TABLE 1. Characteristics of included studies (N = 26) Study (NOS) Study design (sample size) a Risk factor (tool) Chronic cough (tool) Main objectives of the study

Krzyzanowski, 199031

(NOS = 7)

Cracow cohort

Poland

BS: 1968

First FUP: 1973

Second FUP: 1981

Age 19–60 yrs

(N = 2730, M = 1132)

Occupational exposure to dust, variable temperature CC: cough ≥3 mo for 2 yrs; CB: cough with phlegm ≥3 mo for 2 yrs. Incident case as reported at first or second FUP (IA-Q) Incidence rates of chronic respiratory symptoms and related occupational factors

Krzyzanowski, 199232

(NOS = 7)

FUP: 1981

Age 19–70 yrs

(N = 3082, M = 1264)

Smoking

Relationships of the change of respiratory symptoms to age, gender and smoking, compared in two cohorts

Tucson cohort

United States

BS: 1972

FUP: 1985

Age 19–70 yrs

(N = 1452, M = 613)

Silva, 200433

(NOS = 7)

Age ≥ 20 yrs; 12 FUPs until 1993

(N = 3099, M = 1405)

Asthma

Smoking

Atopy (skin prick test)

CB: cough with phlegm ≥3 mo for 2 yrs (P-Q) Association between doctor-diagnosed asthma and subsequent development of COPD

Eagan, 200234

(NOS = 7)

Norway

Age 15–70 yrs

BS: 1985

FUP: 1996–1997

N = 2819, M = 1352 Smoking CC: cough ≥3 mo for 1 yr (P-Q) Incidence rates for respiratory symptoms and asthma and their associations with sex, age, smoking and educational level

Eagan, 200435

(NOS = 7)

Education as indicator for SES

Skorge, 200936

(NOS = 7)

N = 2312, M = 1123, F = 1189 Occupational exposure to dust, gas or fumes Occupational exposure on incident adult asthma and respiratory symptoms

Guerra, 200537

(NOS = 6)

Pisa cohort

Italy

Age ≥ 15 yrs

BS: 1985–1988; FUP: 1991–1992

(N = 1670, M = 592)

Rhinitis

Asthma

Smoking

Occupation

CC: cough ≥3 mo for 2 yrs (IA-Q)

1. Identify risk factors for cough

2. Rhinitis as independent risk factor for occasional and CC

Maio, 201938

(NOS = 7)

Age ≥ 20 yrs

BS: 1991–1992; FUP: 2009–2011

(N = 970, M = 426)

Smoking

Occupation

Vehicular traffic exposure

Usual cough apart from common colds (IA-Q) Cumulative incidence of respiratory symptoms related to smoking, occupation and environment

Hu, 201629

(NOS = 5)

China. Retrospective study. Age ≥ 35 yrs in 2009 long-term (≥3 yrs) residents (N = 1003, M = 312) Air pollution (home–road distance) CC: cough ≥8 wks in the past year (Questionnaire) Home–road distance and lung functions, airway inflammation markers, prevalence of respiratory symptoms and diseases Ellison-Loschmann, 200739 (NOS = 7)

BS: 1991–1993; FUP: 1998–2002; 2010–2012

Age 20–44 yrs

ECRHS 13 countriesb (N = 6455, M = 3412)

SES CB: cough with phlegm ≥3 mo last year (IA-Q) Prevalence and incidence of asthma and CB in relation to SES

Lytras, 201940

(NOS = 8)

ECRHS 15 countriesb (N = 8794, M = 4168)

Smoking

Occupational exposure

Effect of occupational exposures on CB incidence CC: cough ≥3 mo/yr

Butler, 200441

(NOS = 8)

Singapore

Age 45–74 yrs

BS: 1993–1998

FUP: 1999–2002

(N = 49,140, M = 20,786)

Diet (isoflavones, fruit, energy, vitamins, soya) CC: cough ≥3 mo for 2 yrs; CB: CC and phlegm (IA-Q) Fruits, vegetables, decrease risk of developing cough and phlegm

LeVan, 200642

(NOS = 7)

FUP: 1999–2004

(N = 45,104, M = 18,319)

Occupational exposure to dust, vapour, smoke Occupational risk for asthma, cough and phlegm

Baik, 200848

(NOS = 8)

South Korea. Age 40–69 yrs; BS: 2001; FUP: 2003–2005 (N = 4270, M = 2220) Snoring (IA-Q) CB: cough with phlegm ≥3 mo for 2 (IA-Q) Effect of snoring on the development of CB

Holm, 201443

(NOS = 7)

Sweden; age 20–60 yrs. BS: 1993; FUP: 2003 (N = 11,148, M = 5230) Smoking. Asthma. Atopy (hay fever, atopic dermatitis) (P-Q) CB: cough with phlegm ≥3 mo for 2 yrs (P-Q). Onset age asked in 2003 to confirm incident cases Incidence rate of CB in relation to smoking, age, sex, atopy and asthma

Brutsche, 200645

(NOS = 7)

Switzerland. Age 18–60 years. BS: 1991; FUP: 2001 (N = 4855, M = 2573) Bronchial hyper-responsiveness to methacholine CC: cough ≥3 mo for 2 yrs; CB: CC+ phlegm (IA-Q) BHR as risk factors of asthma COPD, respiratory symptoms

Schikowski, 201044

(NOS = 6)

Germany. Females only; age = 45

BS: 1985–1994; FUP: 2006 (N = 2116)

Smoking. Air pollution (PM10, NO2)

(local monitor station data based on home address)

CC: frequent cough without phlegm production. CB: self-reported and physician diagnosed. CC with phlegm (P-Q) Whether air pollution reduction attenuates prevalence of respiratory symptoms

Chhabra, 200130

(NOS = 6)

India. Age ≥ 18 yrs in 1999 and resided in the area for ≥10 yrs (N = 4141, M = 2344) Smoking. Retrospective air pollution data. SES (IA-Q) CC: cough ≥3 mo for 2 yrs (IA-Q) Role of ambient air pollution in chronic respiratory morbidity Mirabelli, 2012 (NOS = 8)52 United States. Age 45–64 yrs; BS: 1980. FUP: 1983 (N = 8967, M = 3949) Occupation (IA-Q) CC: cough ≥4–6 times/day, ≥4 days/wk (IA-Q) Incidence of CC, wheezing, lung function and occupational risks

Xu, 199746

(NOS = 7)

The Netherlands. Age 15–45 yrs

BS: 1965–1969; FUP: every 3 yrs until 1990 (N = 2684, M = 1482)

Airway responsiveness (histamine threshold test) CC: cough ≥3 mo per year. CB: cough with phlegm ≥3 wks for 3 yrs (IA-Q) Increased airway responsiveness and remission of chronic respiratory symptoms

Kagamimo, 199647

(NOS = 6)

Japan. Three cohorts during 1972–1990. Age 6–14 yrs (N = 1796, M = 1003)

Skin prick test

Persistent cough: usually cough in winter (P-Q)c Predictive value of SPT at childhood for future respiratory symptoms

Hedlund, 200649

(NOS = 7)

Sweden. Age 45, 50, 65 yrs. BS: 1985–1986; FUP: 1996 (N = 4754, M = 2341) SES. Smoking. Family history of asthma (P-Q) CB: cough with phlegm ≥3 mo for 2 yrs SES, incidence of asthma and respiratory symptoms

Terho, 199553

(NOS = 7)

Finland. Twin study. Age ≥ 18 yrs. BS: 1975. FUP: 1981 (N = 17,134, M = 9221) Smoking. Farmers allergic rhinitis or allergic dermatitis CB: cough with phlegm ≥3 mo/yr (P-Q) Atopy, smoking and living in farm environment on the development of CB

Arinze, 202054

(NOS = 8)

The Netherlands. Age ≥ 45 yrs

BS: 2009–2014; FUP: 2009–2016

(N = 7141, M = 2984)

Smoking. BMI. ACE inhibitor. COPD, GORD, asthma, OAD, CRS

CC: daily cough ≥3 mo for 2 yrs (IA-Q)

Unexplained CC defined as CC without identifiable risk factors (IA-Q)

Period prevalence, incidence and risk factors of CC

Arinze, 202150

(NOS = 8)

BS: 2002–2008; FUP: 2006–2014

(N = 9824, M = 4549)

Chronic pain, clinically relevant depressive symptoms Bi-directional associations between CC and chronic pain

Wang, 202151

(NOS = 7)

Sweden. From birth to 24 years. BS: 1994–1996 (N = 2519, M = 1156) Early-life exposure and environment from 2 m to 16 yrs (Parents P-Q) CB: usually cough and bring up mucus waking up in the morning during winter (P-Q) Prevalence of CB and early-life risk factors including environmental exposures Abbreviations: ACE, angiotensin-converting enzyme; BHR, bronchial hyper-hyperresponsiveness; BS, baseline study; CB, chronic bronchitis; CC, chronic cough; COPD, chronic obstructive pulmonary disease; CRS, chronic rhinosinusitis; F, females; FUP, follow-up study; GORD, gastro-oesophageal reflux disease; IA-Q, interviewer administratered questionnaire; M, males; mo, month(s); NOS, Newcastle–Ottawa Scale scores; OAD, obstructive airway disease; P-Q, postal questionnaires; PM, particulate matter; SES, socioeconomic status; SPT, skin prick test; yr(s), year(s); wk, week(s). aCombined rows indicated multiple published articles from the same cohort or multiple cohorts reported by the same article. b The countries in the ECRHS study are Australia, Belgium, Denmark, Estonia, France, Germany, Iceland, Italy, Norway, New Zealand, Spain, Sweden, Switzerland, UK and the United States; Estonia and France were excluded from the occupational analysis due to no incident CB cases reported during follow-up. c Baseline questionnaires were filled by parents and confirmed by interviewing the children.

The 26 articles were heterogeneous in terms of study population, definition of chronic cough and potential risk factors, as well as in their statistical analyses (Table 1). The study populations were from 24 countries, mostly Europe and the United States, but five studies (six articles) were from Asian countries.29, 30, 41, 42, 47, 48 Notably, articles from Europe and the United States mostly focused on smoking, occupational exposures and asthma, while articles from Asia mostly focused on air pollution and personal factors (i.e., diet and snoring). Cohorts were generally large (median: 3099; Q1 = 1796, Q3 = 8749), with a long follow-up period (median: 11 years; Q1 = 7 years, Q3 = 13 years). Participants were aged from 15 to over 75 years at baseline, except for two articles,47, 51 which measured risk factors in childhood for adult chronic cough.

Quality assessment

Using the NOS, overall quality was found to be good with 22 out of 26 articles graded ≥7 (out of 9) and four articles graded 6 (Table S6 in the Supporting Information). All articles reported results for self-reported cough, while only one reported physician-diagnosed chronic bronchitis (in addition to self-reported chronic cough as a separate outcome).44 Ten articles31-37, 43, 49, 53 measured exposures through postal surveys only, and nine articles29, 30, 37-39, 44-47 reported a follow-up rate of less than 60%. Six articles29, 30, 42, 44, 47, 51 did not exclude people with chronic cough at baseline. Confounding was considered in all articles and adjusted effect estimates were considered for meta-analysis. However, the adjusted confounders were often heterogenous across articles.

Risk factors for adult chronic cough

Overall, 15 domains of potential risk factors were extracted from the 26 eligible articles, including asthma, atopy and allergy, smoking, occupation, air pollution, socioeconomic status (SES) or education, early-life environmental exposure, diet, snoring, obesity, chronic rhinosinusitis (CRS), COPD, GORD, obstructive airway disease (OAD), chronic pain and clinically relevant depressive symptoms. These factors are summarized in Figure 1. Three risk factors, smoking, education/SES and asthma, met the criteria to be included in the meta-analyses.

List of potential risk factors for chronic cough assessed in eligible studies. Studies were grouped by risk factors, regardless of their measurements and definitions. The size of the bubbles and numbers in parentheses represent the numbers of studies. Hollow bubbles or dots represent factors only assessed by one study but were put under the ‘consistent findings’ column for readability. ‘Inconsistent finings’ refers only to directions of associations, represented by colours of bubbles. *Obesity defined as BMI ≥ 30 kg/m2. COPD, chronic obstructive pulmonary disease; CRS, chronic rhinosinusitis; GORD, gastro-oesophageal reflux disease; OAD, obstructive airway disease including asthma, COPD and asthma–COPD overlap; VGDF, vapour, gas, dust and fumes

Asthma, atopy or allergic disease

Nine articles33, 37, 43, 45, 46, 49-51, 53, 54 measured asthma or asthmatic features (i.e., clinical disease, airway hyper-responsiveness, childhood asthma or family history of asthma) and all found some evidence for positive associations between these factors and chronic cough (Table S7 in the Supporting Information). Our pooled effect estimate showed associations between doctor-diagnosed asthma and chronic cough with minimal heterogeneity (pooled adjusted OR [aOR] = 3.01; 95% CI: 2.33–3.70; I2 = 0%; number of articles (N) = 3), as well as airway hyper-responsiveness and chronic cough (aOR = 2.11; 95% CI: 1.35–2.87; I2 = 18.6%; N = 2). Two articles also found that the HRs of chronic bronchitis increased for people with asthma.33, 43 (Figure 2). Interestingly, one article49 found that family history increased the risk of chronic cough in middle-aged, while another article51 found no association between parental asthma at age 2 months and chronic bronchitis in early 20s. Different ages (for both participants and their families), family members and cough definitions may all contribute to the inconsistency. One article that defined atopy as atopic dermatitis or hay fever found no associations between atopy and chronic cough,43 while two articles found rhinitis and/or allergic dermatitis associated with chronic cough.37, 53 There was no evidence of a relationship between skin prick tests33, 47 or serum IgE levels33 and chronic cough.

Associations between asthmatic features and chronic cough. Fixed-effect models were used. ORs (in normal scale) of asthma (or bronchial hyper-responsiveness) and chronic cough (or chronic bronchitis) were presented. Estimates from ‘results not pooled’ were only shown for comparison without been meta-analysed. CB, chronic bronchitis

Smoking

Compared with never smoking, four articles found that persistent smoking was consistently associated with higher odds of chronic cough (pooled aOR = 1.81; 95% CI: 1.36–2.26; I2 = 57%; N = 3) (Figure 3, Table S8A in the Supporting Information).34, 38, 49 Three articles that investigated associations between past smoking (quitting prior to baseline), remitted smoking (smoker at baseline but quit during follow-up) and chronic cough found no association,38 a positive association49 and a negative association,34 respectively. The substantial heterogeneity (remitted smoking: I2 = 76%; past smoking: I2 = 83%) may be explained by different study designs and populations (Figure 3). Associations between incident smoking and chronic cough were also inconsistent (pooled aOR = 1.28; 95% CI: 0.65–1.90; I2 = 64%; N = 2),34, 38 but small number of events (only 25 participants were incident smokers) may explain some of this inconsistency (Table S8A in the Supporting Information).38

Associations between smoking and chronic cough by different smoking status. All reference groups are never smoking; past smoking is defined as quit smoking before baseline (ex-smoking at baseline); remitted smoking is defined as quit until follow-up (ex-smoking at baseline and at follow-up); fixed-effect models were used

Six articles investigated the association between baseline smoking status and subsequent chronic cough without accounting for any change in smoking status at the time of follow-up (Table S8B in the Supporting Information). All found some evidence for a positive association between current smoking at baseline and chronic cough (pooled adjusted RR [aRR] = 1.97; 95% CI: 1.68–2.27; I2 = 64%; N = 3).33,