Associations of early life and childhood risk factors with obstructive sleep apnoea in middle‐age

Abstract Background and Objective Early‐life risk factors for obstructive sleep apnoea (OSA) are poorly described, yet this knowledge may be critical to inform preventive strategies. We conducted the first study to investigate the association between early‐life risk factors and OSA in middle‐aged adults. Methods Data were from population‐based Tasmanian Longitudinal Health Study cohort (n = 3550) followed from 1st to 6th decades of life. Potentially relevant childhood exposures were available from a parent‐completed survey at age 7‐years, along with previously characterized risk factor profiles. Information on the primary outcome, probable OSA (based on a STOP‐Bang questionnaire cut‐off ≥5), were collected when participants were 53 years old. Associations were examined using logistic regression adjusting for potential confounders. Analyses were repeated using the Berlin questionnaire. Results Maternal asthma (OR = 1.5; 95% CI 1.1–2.0), maternal smoking (OR = 1.2; 1.05, 1.5), childhood pleurisy/pneumonia (OR = 1.3; 1.04, 1.7) and frequent bronchitis (OR = 1.2; 1.01, 1.5) were associated with probable OSA. The risk‐factor profiles of ‘parental smoking’ and ‘frequent asthma and bronchitis’ were also associated with probable OSA (OR = 1.3; 1.01, 1.6 and OR = 1.3; 1.01–1.9, respectively). Similar associations were found for Berlin questionnaire‐defined OSA. Conclusions We found novel temporal associations of maternal asthma, parental smoking and frequent lower respiratory tract infections before the age of 7 years with adult OSA. While determination of their pathophysiological and any causal pathways require further research, these may be useful to flag the risk of OSA within clinical practice and create awareness and vigilance among at‐risk groups.


INTRODUCTION
Obstructive sleep apnoea (OSA) is highly prevalent 1 affecting a billion adults worldwide 2 but remains underdiagnosed and under-recognized [1][2][3] with over half of those with OSA being asymptomatic or not recognizing symptoms. 4This high burden of undiagnosed OSA is concerning given its links with chronic cardiovascular disease, 5 stroke, 5 diabetes mellitus, 5 depression, 6 chronic obstructive pulmonary disease (COPD), 7 asthma, 8 cancer 9 and mortality. 9Early diagnosis and treatment and primary preventive interventions could provide health and economic benefits to individuals and communities 10 but are constrained by limited information available on longitudinal predictors of OSA.
To date, the known risk factors for OSA in adults include obesity (BMI >30 kg/m 2 ), 5,11 central body fat distribution, 11,12 large neck circumference, 11 male sex, 5 older age, 5 variant anatomy that causes crowding of the oropharyngeal and/or velopharyngeal region and collapsible and/or narrow upper airways, 5 and genetic susceptibility as evident by familial inheritance 13 of OSA and influence of ethnicity. 14Some adult OSA may have origins in the childhood. 15Anatomical risk factors for paediatric OSA could remain into adulthood even after therapeutic interventions. 16,17][20] Investigation of the association of childhood risk factors with OSA in adults is ideally performed using prospective data available from childhood to adult life, but the paucity of such data precludes meaningful conclusions about such associations.This could be important knowledge needed to predict, screen for and diagnose OSA early in adults to provide them with beneficial interventions.Such knowledge could also trigger further research into the lifelong evolution and relevant pathophysiology of OSA.We aimed to bridge this gap in knowledge by assessing the associations of childhood health risk factors and risk factor profiles with OSA in middle-age, which could provide new insights into delineating causal mechanisms and pathophysiology of OSA in adults.

Study design and participants
We used data from the Tasmanian Longitudinal Health Study (TAHS), 21 the largest and longest-running respiratory health cohort study in the world.In 1968, TAHS recruited all children born in 1961 (aged 7-years) and attending schools in Tasmania (probands; n = 8583) comprising 99% of the target population.They were followed up periodically, at least once a decade.Setting-up of TAHS and subsequent follow-ups have been described previously. 22,23All surviving and contactable probands (n = 6128) were invited to participate in a survey and a clinical assessment at the age of 51-54 years (between 2012 and 2016) and 3609 (42% of the original cohort) responded.

Parental and early-life risk factors
The potential risk factors for OSA that we investigated included currently known risk factors for childhood OSA 24,25 and general respiratory risk factors that we previously identified. 21Of these, the information on maternal and paternal asthma, maternal and paternal smoking, mode of feeding during the first 3 months or life (breast or bottle fed), doctordiagnosed pneumonia or pleurisy, food or drug allergies, hives, eczema, attacks of hay-fever, frequency of bronchitis, frequency of asthma, history of tonsillectomy and bodyweight at age 7 years was gathered from the survey questions completed by the parents in 1968.Information on preterm (prematurity), birthweight and small for gestational age were extracted from hospital records.Late preterm was defined as born between 34 and 36 weeks + 6 days and moderate or very preterm were born before 34 weeks.Small for gestational age was birthweight <10th percentile for the duration of gestation.Normal birthweight was 2.5 to <4.0 kg.Childhood weight categories were defined using age-sex-specific reference points. 26,27Details of the survey questions that were used and definitions of variables are given in Appendix S1 in the Supporting Information.

Risk factor profiles
We previously considered 14 early life variables for a latent class analysis (LCA) and generated six risk profiles for general respiratory conditions. 21These were: (1) unexposed or least exposed to risk factors (the reference group), (2) parental smoking, (3) allergy, (4) infrequent asthma and bronchitis, (5) frequent asthma and bronchitis and (6) frequent asthma, bronchitis and allergy.Further details of the LCA and risk-factor profiles are given in Appendix S2 in the Supporting Information.

SUMMARY AT A GLANCE
This study provides the first known evidence for individual and profiled early-life and childhood risk factors for OSA in adults.It shows that early exposures to smoking and lower respiratory tract infections could be risks for adult OSA, which may stimulate future research and help flag future risk of OSA.

Probable OSA
During the 53-year follow-up, the probands completed a self-administered survey that included the STOP-Bang 28 (Snoring, Tiredness, Observed apnoea, high blood Pressure, BMI, Age, Neck circumference and Gender) and Berlin 29 questionnaires (details in Appendix S3 in the Supporting Information) that detect probable OSA.These two questionnaires have standard cut-off scores (≥3 for STOP-Bang 28 and ≥2 for Berlin 29 ) but we found in our recent validation study in the same cohort 30 that ≥5 was the optimum cut-off score for STOP-Bang, which was used in this analysis.The standard cut-off score of ≥2 remained the optimal cut-off score for the Berlin questionnaire and was used in this analysis.

Statistical analysis
We reported descriptive data as numbers and percentages or means and SDs.Any differences in the relevant characteristics between those who did or did not respond to the OSArelated questionnaires in the cohort were reported using the chi-squared test.Logistic regression models were used to examine the associations of individual childhood risk factors and risk factor profiles with probable OSA.In these analyses, probable OSA was primarily reported using STOP-Bang questionnaire, but all models were also refitted using Berlin questionnaire as a sensitivity analysis and reported in Tables S2 and S3 in the Supporting Information.Minimum sets of confounders developed using a directed acyclic graph (Figure S1 in the Supporting Information) were used in modelling.The results were reported as odds  ratios (ORs) with 95% CIs.For risk factors and risk factor profiles that showed statistically significant associations with OSA, we performed relevant mediation analyses. 31The Stata/SE 17.0 software 32 was used for statistical analysis.

RESULTS
At the time of baseline data collection, the mean age ± SD of the probands was 6.5 ± 0.3 years (range 6.0-7.0 years).Other baseline characteristics are given in Table 1 and Table S1 in the Supporting Information.Of those who were invited to participate in the sixth decade follow-up (n = 6128), a total of 3609 (59%) responded to survey questions.Information on childhood risk factors was available for 98% of the participants who responded to the STOP-Bang questionnaire.The mean age ± SD at follow-up was 53.0 ± 1.0 (range 50.8-55.6years) and 48.9% were male.In the TAHS cohort (n = 8583), those who had OSA-related information (i.e., those who participated in the sixth decade follow-up) were more likely to be born preterm, have been bottle-fed in the first 3 months of life, had never experienced hives and were less exposed to parental smoking, but more likely to have childhood eczema, hay-fever, frequent bronchitis and history of tonsillectomy during childhood (Table S1 in the Supporting Information).Of the participants of sixth decade follow-up, those who had OSA-risk were more likely to have a history of maternal asthma, maternal smoking, frequent asthma by age 7 years and a higher frequency of pleurisy or pneumonia and frequent bronchitis compared with those who did not have an OSA-risk, but less likely to have history of being solely fed on breast milk (Table 1).
6][37] We found no significant mediation for any of the variables (data not shown).

DISCUSSION
This is the first study to assess the role of prenatal and early childhood factors on OSA in middle-age.We found that maternal asthma and smoking, lower respiratory tract infections and the risk-factor profiles of parental smoking and frequent asthma and bronchitis in childhood were associated with a probable OSA in middle-age.
Although these associations are longitudinal, it is difficult to determine if they are causal or mere predictors of OSA given the current understanding of the pathophysiology of OSA.There are four widely accepted main pathophysiological pathways: 38 an anatomically or functionally collapsible upper airway; poor upper airway dilator muscle responsiveness; high loop-gain of the respiratory system causing respiratory control instability; and low respiratory arousal threshold.Nevertheless, it is difficult to directly link any of the risk factors that we detected to these pathophysiological pathways.For example, this current knowledge makes it difficult to theorize how maternal asthma would cause OSA in adult offspring.However, as OSA symptoms might be misconstrued as nocturnal asthma, 39 at least some of the maternal asthma could be maternal OSA and may indicate a familial inheritance of OSA as previously shown. 13urthermore, maternal asthma could also suggest abnormal airways, a trait the offspring could inherit and be associated with adult OSA.However, the pathophysiology of this association should be explored in future research and the studies of OSA epidemiology must explore maternal asthma as an alternative pathway to OSA.Similarly, the pathophysiology of the association between early exposure to smoking and subsequent OSA remains to be sufficiently characterized.Yet, given that passive smoking has been linked to childhood OSA in at least some studies, 24,40 it is possible that this association could, at least partially, be due to childhood OSA persisting to the adult life.Although obesity in adults was associated with OSA, 41 BMI at the age of 7 years was not associated with OSA in adults.We have previously shown that the BMI from childhood to middle age takes five different trajectories, 42 and if the effect of BMI on OSA was immediate rather than longlasting as previously shown, 43 this would explain the lack of association between childhood BMI and OSA in adults.Similar to childhood obesity, other risk factors for childhood OSA 20 such as prematurity and feeding mode were not associated with OSA in adults suggesting different pathophysiological pathways for OSA in children and adults.
Recurrent respiratory tract infections are associated with childhood OSA, 44 and so it is likely that at least some of these adults with OSA had childhood OSA which continued into adult life.Childhood lower respiratory tract infections are associated with restrictive and obstructive lung diseases 45 which are also associated with OSA. 33,46However, our mediation analysis showed no significant mediation of these observed associations via asthma or COPD in adult life.In adults, an association of chronic bronchitis with OSA 47 and OSA symptoms 48 had been shown.However, we demonstrated that childhood bronchitis is not associated with chronic bronchitis in adults 49 so our finding of association of bronchitis in children with OSA in adults is novel.Frequent lower respiratory tract infections including bronchitis also retard the growth of the respiratory system, 50 and any resulting anatomical or functional changes that narrow the airways and/or make them collapsible would likely increase the risk of OSA. 38On the other hand, the likelihood of an unobserved latent factor driving both respiratory tract infections during childhood and OSA in the adult life cannot be excluded.
Our study has both strengths and limitations.The main advantage in our prospective population-based design is that it has eliminated potential recall bias and ensured wide generalizability.By using the Berlin questionnaire in sensitivity analyses, we have enabled consistent and triangulated observations which strengthen the validity of our findings.These two questionnaires have different sensitivity and specificity levels, 30 yet showed similar results for the associations.However, despite these data coming from a cohort study, we cannot attribute causal inferences to our findings as OSA was detected from cross-sectional data rather than true incident data due to lack of data on childhood OSA.Therefore, the detected associations would best serve to generate hypotheses, which must be tested in further studies.
The main limitation of our study was the reliance on STOP-Bang ≥5 to define OSA instead of sleep studies.STOP-Bang ≥5 has high specificity but less sensitivity 30 and therefore, we may have undercounted OSA.Fortunately, any such misclassification is likely to be non-differential and if so, would have underestimated the observed associations and may also be a reason why some risk-factors were detected with marginal statistical significance.It could also mean some true associations between potential risk factors and OSA went undetected in our study.Sleep studies may detect stronger associations but would be logistically difficult with the numbers required for epidemiological studies.Any studies with polysomnography would likely have much smaller numbers to analyse and so lack statistical power.Missing childhood data for some exposure variables in our study may also have led to reduced statistical power.With multiple modelling using individual variables and likely risk factor profiles, the probability of us detecting a spurious association increased and therefore, these results must be interpreted with caution.
Despite these limitations and the inability to draw causal inferences, this first known study to investigate the role of both individual and profiles of childhood risk factors reinforced the suggestions of childhood origins of adult OSA [15][16][17] and provided new insights into modifiable early life risk factors for OSA, which may stimulate future research.There were strong signals that early exposures to smoking and lower respiratory tract infections could be risks for adult OSA, which could be used in general practice as a part of the routine clinical check-ups to identify those who may be at risk of developing OSA.This knowledge could also be used in a larger scale in public health education to create population awareness, enabling those at risk to be vigilant of OSA which is likely to help detect any OSA early.
In conclusion, maternal asthma, parental smoking and frequent lower respiratory tract infections in childhood were associated with a probable OSA in middle-age.These findings are a likely stepping-stone for further research on early life risk factors for OSA in adults and provide a platform to investigate plausible pathophysiological mechanisms.

a
Probable OSA = STOP-Bang score ≥5.b No adjustment needed as per the causal model (Figure S1 in the Supporting Information).c Adjusted for maternal asthma.d Adjusted for paternal asthma.eResultsdid not materially change when analysed as a dichotomised variable (full-term vs. preterm).f Adjusted for maternal smoking and paternal smoking.g Adjusted for maternal smoking, paternal smoking, prematurity and small for gestational age.h Adjusted for prematurity, small for gestational age and birthweight.i Results did not materially change when analysed as a dichotomised variable (normal weight vs. overweight or obese).j Adjusted for prematurity, small for gestational age, birthweight and breast/bottle feeding.k Adjusted for maternal asthma, and paternal asthma, maternal smoking, paternal smoking, gender, breast/bottle feeding and BMI at age 7 years.l Adjusted for childhood BMI.
Distribution of childhood factors in those with or without probable OSA a at 53 years of age.
Abbreviation: OSA, obstructive sleep apnoea.a Probable OSA = STOP-Bang score ≥5.b Only those with childhood data were included.EARLY-LIFE RISKS FOR OSA IN ADULTS Association between childhood risk factors and probable OSA a in middle-age.
T A B L E 2Abbreviation: OSA, obstructive sleep apnoea.
T A B L E 3 Association between childhood risk factor-profiles and probable OSA a .
a Probable OSA = STOP-Bang score ≥5.b Adjusted for maternal asthma and paternal asthma as per the causal model (see Figure S1 in the Supporting Information).c No adjustment needed as per the causal model.d Adjusted for maternal smoking, paternal smoking, maternal asthma, paternal asthma, breast/bottle feeding, childhood BMI and gender.EARLY-LIFE RISKS FOR OSA IN ADULTS