Trajectories of cough without a cold in early childhood and associations with atopic diseases

Although children can frequently experience a cough that affects their quality of life, few epidemiological studies have explored cough without a cold during childhood.


| INTRODUC TI ON
Cough is a frequent and non-specific respiratory symptom in children that may considerably reduce their quality of life. 1 In the majority of otherwise healthy children, cough is a symptom related to a self-limiting viral upper respiratory tract infection that resolves within a week. 2 However, children with coughs that are not associated with a respiratory infection but are often triggered by normally innocuous stimuli are commonly seen in paediatric practice. This type of cough is more frequent in children who also wheeze and the presence of the two concomitant symptoms is well documented in asthma.
Regardless of associated asthma, coughs triggered by normally nontussigenic stimuli highly suggest cough hypersensitivity, where dysregulated afferent neural pathways and/or the central processing of the cough are likely mechanisms 3,4 with atopy as one of the aetiological mechanisms. 5 Understanding of the mechanisms behind the transition to chronic cough is still a great challenge. Recently published European Respiratory Society guidelines on the diagnosis and treatment of chronic cough refer to its predominant causes in children. 6 Systematic reviews and studies reveal that cough aetiology, frequency and sensitivity differ in many ways throughout childhood 5,7 and between children and adults, 8,9 so these guidelines recommend exploring the natural history of cough over time through observational cohort studies.
A data-driven approach, based on unsupervised statistical methods such as (LCA), has increasingly been used to explore the natural course of respiratory symptoms. Most studies using this approach focus on asthma, 10,11 and a few focus on allergic rhinitis and atopic sensitization. 12,13 Cough data have been mostly used to determine wheeze and asthma phenotypes. 14 While atopy and Th2 cell-mediated inflammation have been considered one of the aetiological mechanisms of cough hypersensitivity syndrome, no study has explored the relationship between cough and all atopic diseases, including food allergy and atopic dermatitis. In most studies that have investigated respiratory symptoms' trajectories throughout childhood, cough, wheezing and asthma were grouped together.
Thus, cough is always explored as an asthma symptom and no studies focus on cough patterns.
The European prospective birth cohort PASTURE (Protection against Allergy STUdy in Rural Environment) involves children from rural areas and aims to evaluate risk and protective factors for allergic diseases, offering the opportunity to explore trajectories of cough without a cold in childhood and investigate their link with atopic diseases.
This study aims to assess trajectories of cough without a cold in childhood from 1 to 10 years old in the PASTURE cohort and their associations with atopic diseases, including asthma and the farming environment.

| Study design and population
The PASTURE/EFRAIM (Mechanisms of early protective exposures on allergy development) study focuses on a prospective birth cohort involving children born in 2002 and 2003 in rural areas in five European countries (Austria, Finland, France, Germany and Switzerland) to evaluate risk factors and protective factors for allergic diseases. The design of the PASTURE study has been described in detail elsewhere 15 and the inclusion and exclusion criteria of this study, detailed in online supplements, were those of the overall PASTURE study. 16 To briefly summarize, pregnant women were recruited during their third trimester of pregnancy and divided into two groups: women who lived on family-run farms where livestock was kept (farm group) and women from the same rural areas who did not live on a farm (non-farmer group). In total, 1133 children were included in this birth cohort. The study was approved by the local research ethics committee in each country, and written informed consent was obtained from the parents.

| Questionnaires
Questionnaires were self-administered by the parents when the children were 12, 18, 24, 36, 48, 60 and 72 months old and then at 10 years of age. The questionnaires were based on items from the International Study of Asthma and Allergies in Childhood, 17 the Asthma Muti-centre Infants Cohort Study 18 and the American Thoracic Society. 19 At all time-points, parents were asked "How often has your child had a cough without a cold during the last 12 months?" (or during the last 6 months at the 18-and 24-month follow-ups). The possible answer categories were "never," "less than once a month," "once a month" and "at least twice a month." The same question was asked for "cough at night without a cold." When the children reached 2 years of age, parents were also asked "Has your child ever had an attack of cough without a cold caused by one of the following factors: physical exercise, excitation, change of temperature?" Allergy and asthma assessment is needed in case of persistent cough without a cold.
Growing up on a farm seems to have a prot ve effect on acute transient cough and early persistent cough.

G R A P H I C A L A B S T R A C T
Five different patterns of cough trajectories during early childhood have been highlighted. Recurrent cough without a cold, with night cough and triggers should lead to allergy and asthma assessment. Growing up on a farm seems to have a protective effect on acute transient cough and early persistent cough.
The length and characteristics of cough are described in the Appendix S1.
Unremitting wheeze was defined by the prevalence of wheeze without a cold or symptoms between wheezes reported by the parents at least once between 18 months and 10 years of age. Children were considered as having unremitting wheeze if the parents answered "one or more" to the question "How many attacks of wheezing has your child had in the last 12 months apart from a cold?" or "no" to the question "Is your child completely cured (without any respiratory complaints) between these episodes?" Children were defined as having doctor-diagnosed asthma if the parents reported that the child had been diagnosed with asthma by a doctor at least once or if the child had had at least two doctordiagnosed spastic, obstructive or asthmatic bronchitis in questionnaires at age 4, 5, 6 and 10, independent of a diagnosis reported in the first 3 years of life.
Allergic rhinitis was defined by the simultaneous presence of nasal and eye symptoms without a cold (itchy, runny or blocked nose and red, itchy eyes) and/or a doctor-diagnosed allergic rhinitis from 3 to 10 years of age.
Definitions of atopic dermatitis, food allergy, parental history of atopy and specific IgE (sIgE) measurements are provided in the Appendix S1.

| Statistical analysis
A LCA was used to identify subtypes of cough symptoms over time. 20 The variables "cough without a cold" and "cough without a cold at night" were coded in binary variables "never" versus "at least once during the last 12 months" at each visit. Cough triggered by physical exercise, change of temperature and/or excitation was coded in binary variables "no trigger" versus "at least one trigger" at each visit. The three defined variables were incorporated into the LCA model. As we decided to focus on the three items that supported the diagnosis of asthma as described by the Global Initiative for Asthma (GINA), we did not incorporate the length and characteristics of cough without a cold into the LCA model.
Children with data on cough symptom at less than six of the eight visits were excluded (n = 172). Bayesian information criterion (BIC), consistent Akaike information criterion (cAIC), and entropy were used to define the number of classes that best fit the data. 21 BIC was considered as the most reliable fit statistic and bootstrapped likelihood ratio test was performed in case of discordance between two statistical parameters. 22 The probability of an individual belonging to each class was estimated based on conditional probabilities of cough symptoms at each time given a class membership. 23 For sensitivity analyses, a LCA was performed on children with information on cough symptoms at all eight time-points and among the entire PASTURE population.
Multinomial logistic regression was used to investigate the associations between latent class trajectories of cough (outcomes) and the characteristics of the population (exposures). Logistic regression was used to investigate the associations between atopic diseases (outcomes) and the trajectories of cough (exposures). Multivariable models were adjusted for centre, parental history of atopy, gender and farming status, as there were known associations with allergic diseases and the centres used in the study population selection.
Stratified analyses were performed to investigate the associations between respiratory atopic diseases and trajectories of cough according to unremitting wheeze. A data analysis was performed using SAS software version 9.4 (SAS Institute Inc.).  Table S1.

| Study population
Regarding allergic diseases, 815 subjects presented available data for atopic dermatitis, 781 for food allergy, 773 for allergic rhinitis and 775 for doctor-diagnosed asthma. sIgE assays were performed on 521 children at the age of 10.
The point prevalence of cough symptoms during the first 10 years of life is presented in Table 1.

| Selection of the class solution that best fit the data
According to BIC and entropy, the nine-class solution was identified as the best model to fit the data ( Table 2). cAIC was the lowest for the eight-class solution but the bootstrapped likelihood ratio test was significant in favour of the nine-class solution (p = .01). Apart from one large class (n = 576; 59.9%), the eight other classes represented 4.4% (n = 42) to 6.8% (n = 65) of the population. Sensitivity analyses are described in the online supplements. and a cough at night increasing from 1 to 5 years of age, and more than a 60% prevalence rate of a cough at 10 years old. The early persistent trajectory had the highest prevalence of triggers (from 30% to 60%).

| Trajectories of cough without a cold
In all trajectories, most coughing episodes lasted less than a week ( Figure S3). In both persistent trajectories, about 50% of children with a cough had episodes lasting a week or more and about 10% had coughing episodes lasting more than 2 weeks. Apart from the reference, dry cough was predominant ( Figure S4).

| Association of cough trajectories with the characteristics of the study population
Growing up on a farm was inversely associated with the acute transient and early persistent cough trajectories (Table 3). These associations persisted after adjustment for the centre and parental history of atopy. Parental history of atopy was positively associated with the acute transient, late and early persistent trajectories. Gender was not associated with cough trajectories.
TA B L E 1 Point prevalence of cough symptoms up to 10 years of age.
Cough triggered by change of temperature Out of children with cough without a cold Abbreviations: CI, confidence interval; NA, not applicable. a At least once in the last 12 months (6 months at the 1.5-and 2-year follow-up).

| Association of cough trajectories with atopic diseases and sensitization
Unremitting wheeze was associated with all cough trajectories, with the strongest association for the persistent trajectories (Table 4).
Both doctor-diagnosed asthma and allergic rhinitis were positively associated with all cough trajectories, with the strongest associations for early persistent. Regarding pooling asthma or allergic rhinitis diagnosis, 40.5% of children in the early persistent trajectory had neither allergic rhinitis nor asthma (vs. 89.8% in the reference trajectory).
Of all children with doctor-diagnosed asthma, 87% provided information about their age when they were first diagnosed. Among those, 57.5% were first diagnosed before or at the age of three and 8.0% were diagnosed aged six or over. There was no statistical difference in age at the first diagnosis of asthma according to cough trajectories (p = .2225). The cumulative prevalence of asthma according to the different cough trajectories is presented in Figure 2. Most asthma diagnoses were reported by the age of five, even in the late persistent trajectory. However, the early persistent trajectory had the highest proportion of reported asthma diagnoses after the age of five (22.7%).
Food allergy was associated with late and early persistent trajectories (Table 4). Atopic dermatitis was not associated with any cough trajectories. sIgE sensitization to seasonal and perennial aeroallergens was associated with the late persistent trajectory.

| Stratified analyses according to unremitting wheeze
In children with unremitting wheeze (    Prevalence of other atopic diseases and sensitization according to unremitting wheeze are presented in Table S2.

| DISCUSS ION
To the best of our knowledge, this study is the first to assess trajectories of cough without a cold during the first 10 years of life using statistical methods without a priori assumptions. Apart from children who never or infrequently suffered from a cough in the absence TA B L E 4 Association between the latent class trajectories of cough without a cold and unremitting wheeze, atopic diseases up to 10 years of age (asthma, allergic rhinitis, atopic dermatitis and food allergy) and sensitization to food and inhalant allergens at 10 years of age (>0.7 IU/ml).  Regarding the point-prevalence of a cough without a cold during the first 10 years of life, it ranged from 12% to 21%, with the highest prevalence at 1 year of age. That might be explained by physiological gastro-oesophageal reflux disease in the early infancy, as this aetiology of cough apart from colds has been described as frequent. 24 In our cohort, dry cough was predominant except in the reference trajectory. A predominant wet cough in this trajectory suggests a post-viral cough. The prevalence of a cough without a cold, a cough at night and cough triggers were lower than in the Leicestershire cohort (from 34% to 55% from a cough without a cold, 20% to 31% for a night cough and 18% to 26% for triggers). 25 This difference could  be due to population characteristics: rural in the PASTURE cohort and mostly urban in the Leicestershire cohort. The prevalence of a night cough in the Leicestershire cohort was even higher than in the PARIS birth cohort (14%-18%), including children born in the area of Paris. 26 The comparison with other cohorts highlights the originality of our prospective study, which includes several repeated questions about coughing throughout the first 10 years of life. In the PARIS birth cohort, 26  The association of all the cough trajectories with doctordiagnosed asthma and allergic rhinitis demonstrates the close relationship between cough, asthma and allergic rhinitis. The association between these atopic diseases, parental history of atopy and acute transient trajectories is contradictory to previous studies. 14,24,25 Even in children without unremitting wheeze, the acute transient trajectory was associated with doctor-diagnosed asthma and allergic rhinitis. These results suggest that a cough without a cold, even acute, is not physiological and could be related to an allergic hypersensitivity.

LCA trajectories of cough without a cold
The highest prevalence of doctor-diagnosed asthma and allergic rhinitis was found in children with the early persistent trajectory. About 40% of children presenting this cough trajectory did not have a diagnosis of asthma or allergic rhinitis at 10 years of age.
These results confirm that an isolated cough without a cold, even at night, is not always asthma. 31 In our study, doctor-diagnosed asthma was mostly reported before the age of three. At this young age, an asthma diagnosis is clinical based on the child's history of wheezing, 32

| CON CLUS ION
We investigated trajectories of cough without a cold from one to 10 years old in a rural prospective birth cohort non selected for risk of atopy. In this population, we identified nine different cough trajectories. The late and the early persistent trajectories, which represent 9.2% of children in our cohort, had the strongest association with asthma and allergic rhinitis and were also associated with food allergy and parental atopy. In clinical practice, these results allow to conclude that children having recurrent cough without a cold and with night cough and triggers should benefit from an asthma and allergy assessment. The strong inverse association between farm environment and an early persistent cough deserves further exploration to seek the prevention of severe cough phenotypes.

AUTH O R CO NTR I B UTI O N S
A.D-C. was involved in acquisition of data, was responsible for sta-

ACK N OWLED G EM ENTS
We thank all the fieldworkers and other PASTURE/EFRAIM team members. We would like to dedicate this manuscript to Jean-Charles Dalphin, one of the founders of the PASTURE study, who passed away in 2019. We are very grateful to him for his investment in the PASTURE cohort from the outset.

FU N D I N G I N FO R M ATI O N
The PASTURE study was supported by the European Commission

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.