Mild‐to‐moderate obstructive sleep apnea and mortality risk in a general population sample: The modifying effect of age and cardiovascular/cerebrovascular comorbidity

About 5.4%–45.7% of the general population has mild‐to‐moderate obstructive sleep apnea (mmOSA), which is highly comorbid with cardiovascular and/or cerebrovascular diseases (CBVD). We examined the association between mmOSA and all‐cause mortality and the modifying effect of age and CBVD. A total of 1681 adults 20–88 years old from the Penn State Adult Cohort (PSAC) (41.9% male) were followed up for 20.1 ± 6.2 years for all‐cause mortality. Mild and moderate OSA were defined as an apnea/hypopnea index (AHI) 5–14.9 and 15–29.9 events/hour, respectively. CBVD was defined as a report of a physician diagnosis or treatment for heart disease and/or stroke. Cox proportional hazards regression models were used to estimate all‐cause mortality adjusted for confounders. All‐cause mortality risk was significantly increased in the mmOSA group in young and middle‐aged adults (<60 years) (HR = 1.59, 95%CI 1.08–2.04) but not in older adults (≥60 years) (HR = 1.05, 95%CI 0.80–1.39). A synergistic effect between mmOSA and CBVD was stronger in those <60 years (HR = 3.82, 95%CI 2.25–6.48 in <60 years vs 1.86 95%CI 1.14–3.04 in ≥60 years). There was an additive effect between moderate OSA and hypertension in <60 but not in those ≥60 years. Mild OSA was associated with all‐cause mortality only in the presence of CBVD. Mortality risk is increased in young and middle‐aged adults with moderate OSA, whereas the mortality risk associated with mild OSA is elevated only, regardless of age, in the presence of comorbid CBVD. AHI cut‐offs warranting treatment of mmOSA may need to be adjusted based on age and comorbidities.

To date, only three studies have examined the association between mild-to-moderate obstructive sleep apnea (mmOSA) with mortality population-and community-based samples.All three studies reported an association between severe OSA and mortality, but no association was observed between mmOSA and all-cause mortality (Marshall et al., 2014;Punjabi et al., 2009;Young et al., 2008).

Cross-sectional and longitudinal studies from the Penn State
Adult Cohort (PSAC) and the Sleep Heart Health Study (SHHS) have shown that OSA is more strongly associated with hypertension, in young and middle-age adults compared with older individuals (Bixler et al., 1998;Bixler et al., 2000;Haas et al., 2005;Newman et al., 2001;Vgontzas et al., 2019).However, the potential of an agemodifying effect on the association between OSA and hypertension has not been either tested or found in other studies (Geovanini et al., 2018;Heinzer et al., 2015;Peppard et al., 2000).Interestingly, two previous studies have reported that in patients with severe OSA, mortality risk is increased in younger but not in older patients (He et al., 1988;Lavie & Lavie, 2009).Cardiovascular and cerebrovascular disorders (CBVD) such as hypertension, heart disease, and stroke are highly comorbid with OSA (Bixler et al., 2000;Cowie et al., 2021).Also, these disorders are independent risk factors of all-cause mortality (Tsao et al., 2022).To our knowledge, no study has examined whether the combination of mmOSA with these cardiometabolic conditions is associated with increased risk for mortality.
The overall aim of this study was to examine the association between mmOSA and all-cause mortality in a random general population sample with a wide age range (20-88 years) over a follow-up period of 20.1 ± 6.2 years.Additional objectives of this study included: (a) examination of the potential modifying effect of age on all-cause mortality, (b) exploration of the impact of mmOSA combined with CBVD and hypertension on all-cause mortality.Based on the above reviewed literature, we hypothesised that mmOSA is associated with increased mortality risk in young and middle-aged versus older individuals and that the risk is increased in the presence of CBVD and hypertension in the same age group.

Participants
The data presented here were collected as part of the PSAC, a population-based study of sleep disorders, which used a 2-phase protocol in order to recruit participants from various age groups (Bixler et al., 1998;Bixler et al., 2001;Bixler et al., 2002;Kish, 1965;Waksberg, 1978).In the first phase of the study, telephone interviews were conducted with 16,583 age-eligible men and women with response rates of 73.5% and 74.1%, respectively.In the second phase, 741 men and 1000 women were randomly selected from the first phase and studied in the sleep laboratory, with response rates of 67.8% and 65.8%, respectively.Detailed descriptions of the sampling procedure have been described elsewhere (Bixler et al., 2000).After giving a complete description of the study to the subjects, written informed consent was obtained.The study protocol study complied with the Declaration of Helsinki and was approved by the institutional review board at the Penn State College of Medicine (Hershey, PA).

Mortality follow-up
Death certificates for deceased individuals as of 31 December 2018 were retrieved from the US Centers for Disease Control and Prevention (Fernandez-Mendoza et al., 2019;Fernandez-Mendoza et al., 2020;Fernandez-Mendoza, He, LaGrotte, et al., 2017;Fernandez-Mendoza, He, Vgontzas, et al., 2017).Participants were linked by Centers for Disease Control and Prevention to death records from the National Death Index for the years 1992-2018, and vital status was determined through a rigorous process of probabilistic matching and death certificate review based on participant Social Security number, full name, date and state of birth, sex, race/ethnicity, state of residence, and marital status (Fernandez-Mendoza et al., 2019;Fernandez-Mendoza et al., 2020;Fernandez-Mendoza, He, LaGrotte, et al., 2017;Fernandez-Mendoza, He, Vgontzas, et al., 2017).The primary underlying cause of death was abstracted for case definition, which was classified using International Classification of Disease, Ninth Revision and Tenth Revision for deaths occurring before 1998 and deaths occurring in 1999 and beyond, respectively (Fernandez-Mendoza et al., 2019;Fernandez-Mendoza et al., 2020;Fernandez-Mendoza, He, LaGrotte, et al., 2017;Fernandez-Mendoza, He, Vgontzas, et al., 2017).Data misclassification was avoided by following standard guidelines and algorithms (Fernandez-Mendoza et al., 2019;Fernandez-Mendoza et al., 2020;Fernandez-Mendoza, He, LaGrotte, et al., 2017;Fernandez-Mendoza, He, Vgontzas, et al., 2017).
Of the 1741 participants, a total of 1077 subjects were alive at follow-up and 664 were deceased.Duration of follow-up was calculated from the time of the baseline evaluation to the date of death for those deceased or to 31 December 2018 for those alive.Average follow-up duration in the entire cohort was 20.1 ± 6.2 years.Our analyses focussed on all-cause mortality.

Sleep laboratory evaluation
All participants were evaluated for one night for 8 h in the sleep laboratory.Apneas and hypopneas were defined based on standard criteria, as reported previously (Bixler et al., 2000).For the purpose of this study, participants with severe OSA (AHI ≥30 events/hour) (n = 60) were excluded from the analyses, and mmOSA was defined as an AHI between 5 and 29.9 events/hour.The reference group was defined as participants with an AHI <5 events/hour.

Other measurements
As part of the physical examination during the subject's sleep laboratory visit at baseline, height (cm), and weight (kg) were measured using a calibrated scale and body mass index was calculated (BMI) [body mass index = mass (kg)/(height(m) 2 )].All subjects completed a standardised questionnaire to assess the presence of sleep disorders, physical health conditions, mental health problems, and substance use.The presence of physical health conditions was defined as a binary variable, including any positive response to a past or current history of a physician diagnosis or treatment for allergies/asthma, anaemia, birth defects, cancer/tumour, colitis, encephalitis, epilepsy, kidney/bladder disorders, migraine, Parkinson's disease, rheumatism, thyroid, or ulcer (Bixler et al., 2002;Fernandez-Mendoza et al., 2019;Fernandez-Mendoza et al., 2020;Saulnier et al., 2022).The presence of mental health conditions was also defined by a past or current history of a physician diagnosis or treatment for depression, suicidal ideation or attempts, loneliness, marital problems, alcohol abuse, or drug abuse (Bixler et al., 2002;Fernandez-Mendoza et al., 2019;Fernandez-Mendoza et al., 2020;Saulnier et al., 2022).Participants' daily consumption of tobacco (number of cigarettes/day), and alcohol (number of drinks/day) as well as demographic characteristics, including age, sex, and race/ethnicity, were also obtained (Fernandez-Mendoza et al., 2019).Hypertension at baseline was defined by a selfreport of receiving antihypertensive medication, history of a hypertension diagnosis and/or systolic and diastolic blood pressure levels ≥140/≥90 mmHg, as described previously (Bixler et al., 2000).Diabetes at baseline was defined as a self-report of receiving treatment for diabetes or having a fasting blood sugar ≥126 mg/dL from blood drawn upon awakening in the sleep laboratory, as described previously (Vgontzas et al., 2009).CBVD was defined by a self-report of a physician diagnosis or treatment of heart disease and/or stroke, as described previously (Fernandez-Mendoza et al., 2019;Fernandez-Mendoza et al., 2020).

Statistical analysis
The independent variable was the presence of mmOSA at baseline and the outcome was all-cause mortality, with age (i.e., <60 vs. ≥60 years) as a potential effect modifier.To quantify the excess risk of all-cause mortality associated with mmOSA compared with the reference group (AHI <5), multivariable-adjusted Cox proportional hazards regression models were used.We evaluated the significance of the interaction term between the independent variable (mmOSA) and log-transformed survival time to test whether the proportional hazards assumption was violated.The p-value for this interaction term was 0.21, which indicates that the proportional hazards assumption was not violated.Thus, the effects of mmOSA on mortality were presented as hazard ratios (HRs) with their 95% confidence intervals (CI), adjusted for race/ethnicity, sex, BMI, smoking, alcohol consumption, education, sleep difficulty, total sleep time, hypertension, diabetes, and other physical and mental health conditions.We further tested the interaction between mmOSA and age to investigate the hypothesised effect modification.Given the significant effect modification we present the associations on the basis of two strata: <60 and ≥60 years.Further, we projected the multivariable adjusted survival functions using Kaplan-Meier curves for the two age groups (Figure 1).These survival curves were based on the aforementioned multivariable-adjusted Cox models, with the average sample characteristics presented in Tables 1 and 2. Finally, we separately tested the association between mild (AHI 5-14) and moderate (AHI 15-29) OSA with all-cause mortality after adjusting for the same confounders.
To explore whether there was a synergistic effect of mmOSA and CBVD or hypertension on mortality, we tested the interaction between mmOSA and CBVD or hypertension.We observed a significant interaction between mmOSA and CBVD suggesting the presence of a synergistic effect.Even though the interaction between mmOSA and hypertension was not statistically significant, we explored the potential additive effect between the two conditions given that hypertension is highly comorbid with OSA (Cowie et al., 2021) and both of them are associated with adverse health outcomes (Tsao et al., 2022).In these analyses, the independent variables consisted of the presence of mmOSA and/or comorbidities (i.e., hypertension or CBVD) at baseline as separate categories.These independent variables were constructed as 4-level categorical variables (e.g., 1: neither mmOSA nor hypertension, 2: mmOSA only, 3: hypertension only, and 4: both mmOSA and hypertension).We also evaluated the potential violations of the proportional hazards assumption.The interaction terms between these secondary independent variables and logtransformed survival time were not statistically significant (mmOSA + CBVD with log-time: p = 0.84; mmOSA + hypertension with logtime: p = 0.31), which also indicates that the proportional hazards assumption was not violated.As a result, we quantified the excess risk of all-cause mortality associated with mmOSA and/or comorbidities compared with the reference group (AHI <5 and no hypertension or CBVD) using Cox proportional hazard models, while adjusting for the same set of covariables included in the analyses above.Similarly, we also tested the interaction between mmOSA and/or comorbidities (i.e., CBVD and hypertension) and age groups (i.e., <60 and ≥60 years) to assess potential age effect modifications.The interaction terms between age groups with mmOSA + CBVD and mmOSA + hypertension were both statistically significant.Therefore, stratified analyses were performed to assess the relationship between these secondary independent variables and mortality in each age group.Finally, since there was an association between mmOSA and mortality only among those who were <60 years, we further examined separately the association between mild and moderate OSA with mortality, stratifying by CBVD or hypertension status (presence or not) in this age group.
In sensitivity analyses, when we further adjusted for treatment for OSA and self-reported BMI changes across time points, the findings were similar.A value of p ≤ 0.05 was used to determine the significance for all analyses.All analyses were conducted with SAS, version 9.4 (SAS Institute, Cary, NC).

RESULTS
By 31 December 2018, out of the 1681 subjects, a total of 1048 subjects were alive at follow-up and 633 were deceased.

Association between mmOSA with all-cause mortality: Age effect modification
A total of 494 individuals with AHI <5 and 139 individuals with AHI 5-29.9 were deceased.Overall, mmOSA was not significantly associated with increased mortality (Table 3).Based on our hypothesis, the interaction between mmOSA and age was statistically significant ( p-interaction = 0.02) and mmOSA was significantly associated with increased mortality in adults aged <60 years, but not in adults aged ≥60 years (Table 3).Further, moderate but not mild OSA was associated with increased mortality.When we examined the association of OSA and mortality risk stratified by severity of apnea and age (interaction of OSA severity and age was significant p = 0.01) the risk was significant in the younger group of <60 years for moderate OSA but not for mild OSA (Table 3).When we repeated the analyses using minimum oxygen saturation instead of AHI the results were similar (Table S1).Finally, there was not a significant difference between males versus females in terms of the association between mmOSA and mortality.

Combined impact of mmOSA and cardiovascular/ cerebrovascular morbidity on mortality risk
Association between mmOSA and CBVD with all-cause mortality A total of 477 individuals without CBVD and 156 individuals with CBVD were deceased.The combination of CBVD with mmOSA had a synergistic effect on mortality ( p-interaction = 0.03), while mmOSA without CBVD was not significantly associated with mortality (Table 4).Furthermore, the interaction between 4 level mmOSA + CBVD with age was statistically significant (p-interaction = 0.05).
Thus, when we stratified by age, the combination of mmOSA plus CBVD had an increased effect on the mortality risk in both age groups, albeit it appeared to be 2-fold stronger in those <60 vs. those ≥60 years (Table 4).Finally, the severity of OSA affected the association of CBVD with mortality risk only in the younger group, <60 years (the interaction of OSA severity, CBVD and age was significant, p = 0.05) Specifically, the association of mild OSA in the presence of CBVD with mortality was significant, whereas the presence of CBVD did not affect the association of moderate OSA with mortality  5).When we repeated the analyses using minimum oxygen saturation instead of AHI the results were similar (Table S2).Finally, there was no significant difference between males versus females in terms of the association between comorbid mmOSA and CBVD and mortality.

Association between mmOSA and hypertension with all-cause mortality
A total of 190 individuals without hypertension and 443 individuals with hypertension were deceased.The combination of mmOSA and hypertension was not associated with an additive effect on mortality (p-interaction = 0.55).However, the interaction between 4 level mmOSA + hypertension with age was statistically significant ( p-inter-action<0.01).Thus, when we stratified by age, we observed an additive effect between mmOSA and hypertension in the group of <60 years but not in those ≥60 years old (Table 6).Finally, the severity of OSA affected the association of hypertension with mortality risk only in the younger group <60 years (the interaction of OSA severity, hypertension and age was significant, p < 0.01).Specifically, the association of moderate OSA in the presence of hypertension with mortality was significant, whereas the presence of hypertension did not affect the association of mild OSA with mortality (Table 7).When we T A B L E 3 Association between mild-to-moderate OSA and all-cause mortality repeated the analyses using minimum oxygen saturation instead of AHI the results were similar (Table S3).Finally, there was no significant difference between males versus females in terms of the association between comorbid mmOSA and hypertension and mortality.

DISCUSSION
This study provides four important clinical findings: (1) mmOSA is an independent risk factor for all-cause mortality in young and middleage adults aged 20-59 years, but not in adults 60 years or older (HR = 1.6 vs 1.1), an association that is present only in moderate, but not mild, OSA; (2) the comorbidity of mmOSA with CBVD had a synergistic effect on mortality risk, being 2-fold stronger in young and middle-aged versus older adults (HR = 3.8 vs 1.9); (3) the mortality risk associated with mild OSA became 2-fold and significant when comorbid with CBVD; and (4) moderate, but not mild, OSA comorbid with hypertension had an additive effect on all-cause mortality only in young and middle-aged adults (HR = 2.4 vs 1.1).These findings can shift the way we currently diagnose and treat this highly prevalent, but poorly understood, sleep-related breathing disorder.
Three previous population/community-based studies reported an association of severe OSA, but not milder forms of OSA, with mortality (Marshall et al., 2014;Punjabi et al., 2009;Young et al., 2008).The Wisconsin Sleep Cohort Study reported an association between severe OSA, but not mmOSA, with all-cause mortality (Young et al., 2008).Similarly, the SHHS reported an association between severe OSA but not between mmOSA with all-cause mortality (Punjabi et al., 2009).Furthermore, the study by Marshall et al. reported an association between moderate-to-severe, but not mild, OSA with all-cause mortality (Marshall et al., 2014).The lack of association between milder forms of OSA with mortality in these prior studies could be explained by either the restricted age range (primarily middle-aged 30-60 years or primarily older 40-79 years) (Marshall et al., 2014;Young et al., 2008), small number of deaths (Young et al., 2008), sample size (Marshall et al., 2014) and/or shorter followup period (mean 8.2 years) (Punjabi et al., 2009).In contrast to these previous studies, we observed a significant mortality risk in mmOSA in young and middle-aged individuals (20-59 years), which can be explained by the inclusion of a wide age range (20-88 years) of adults randomly selected from the general population with a long follow-up period (mean 20 years).
The age modifying effect on the association of OSA with mortality has been reported in one study based on a general population sample and two studies based on clinical samples (He et al., 1988;Lavie & Lavie, 2009;Punjabi et al., 2009).He et al. early on, reported that male patients with an apnea index (AI) >20 events/h had significantly higher mortality than those with an AI <20 events/h, and this association was significant only in patients younger than 50 years (Age 51.6 ± 12.0 years) (He et al., 1988).Analyses of the SHHS reported that severe OSA was associated with increased mortality in middle-aged men (40-69 years) but not older men (70-79 years) (Punjabi et al., 2009).In addition, Lavie et al. suggested  OSA (respiratory disturbance index 20-39.9)(Lavie & Lavie, 2009).The data of our study confirm and expand these previous findings that age modifies the effect of OSA on mortality even in the milder forms of the disorder.
Another important finding of our study is that the comorbidity of mmOSA with CBVD had a synergistic effect on mortality, being 2-fold stronger in those 20-59 years compared with those 60-88 years.This effect was stronger in subjects with mild OSA.Furthermore, moderate, but not mild, OSA when combined with hypertension, had an additive effect on mortality only in young and middle-aged but not in older individuals.Although it is well established that these disorders are independent risk factors of all-cause mortality (Tsao et al., 2022), this is the first study to examine and demonstrate the impact of mmOSA on mortality when combined with milder forms of cardiovascular disease (i.e., hypertension) as well as more severe forms of cardiovascular or cerebrovascular pathology (i.e., heart disease and stroke).
From a mechanistic standpoint, the lack of association between mmOSA and mortality in the elderly has been explained due to delayed ischaemic preconditioning, a state in which repeated sublethal ischaemia confers protection from infarction and further ischaemic insults (Lavie & Lavie, 2006).Other explanations include a "ceiling effect" (age is so detrimental that OSA does not confer any additive effect) and a selection bias, that is, older individuals represent a healthier group of survivors (Jun & Polotsky, 2020).Alternatively, proposed classification (Baveno classification) that the impact of treatment is related more to the presence of symptoms and comorbidities and less on AHI (Randerath et al., 2021).
Of interest is that an increasing amount of data suggests that CPAP treatment may not improve survival in people with OSA and comorbid CBVD (Cowie et al., 2021;da Silva Paulitsch & Zhang, 2019;Pengo et al., 2022;Sánchez-de-la-Torre et al., 2020).
However, in these studies the modifying effect of age on treatment outcome was not examined.
The present study has several strengths.It is a longitudinal study, examining a randomly selected sample from the general population with a wide age range (20-88 years), focussed on mmOSA and using in-laboratory PSG and a detailed clinical history and physical examination that ensured careful control for multiple potential confounders.Some limitations should be considered when interpreting our results.
First, the definition "heart disease" is based on a single non-specific question.It is possible that more specific diagnoses such as atrial fibrillation or heart failure have a stronger synergistic/additive effect on the mortality risk across all ages (Chahal & Somers, 2015;Khayat et al., 2015).Second, some participants with mmOSA may have shifted to the severe category over a 20-year period (Punjabi, 2008).
Third, we included individuals who underwent CPAP treatment for OSA at some point in our main analyses because we could not be certain that their OSA was being treated optimally.However, when we further adjusted for OSA treatment, findings did not change and remained significant.Additionally, when we repeated the analysis after excluding participants with OSA treatment findings remained very similar and significant (all p < 0.05).Fourth, the group of adults <60 years with comorbid CBVD and mild OSA consisted of 16 individuals.Thus, the reported 2.2-fold risk of this group with all-cause mortality should be interpreted with caution due to the small size of the group.Finally, OSA when combined with excessive daytime sleepiness might influence the association between OSA and all-cause mortality.
However, we could not test this hypothesis due to the small group that reported moderate/severe sleepiness.In conclusion, mmOSA is associated with a 60% increased risk of all-cause mortality only in young and middle-aged adults, but not older adults.Furthermore, the combined impact of mmOSA with heart diseases and/or stroke on mortality risk is 4-fold in young and middle-aged adults and only 2-fold in older adults.The mortality risk for mild OSA is significantly elevated only when comorbid with heart diseases and/or stroke whereas the comorbidity of moderate OSA and hypertension increases the mortality risk by 2.4-fold in young and middle-age individuals, but not in older adults (1.1-fold).It appears that AHI cut-offs warranting treatment of mmOSA should be adjusted based on age, the presence of comorbidities and the underlying endotype(s).
Adjusted for race/ethnicity, sex, BMI, smoking, alcohol consumption, education, sleep difficulty, total sleep time, hypertension, diabetes, and other physical and mental health conditions.The interaction between mild-to-moderate OSA and age groups (<60 vs. ≥60 years) was statistically significant ( p = 0.02).OSA = AHI ≥5 events per hour of sleep.Association between mild-to-moderate OSA and comorbid CBVD on all-cause mortality Note: Adjusted for race/ethnicity, sex, BMI, smoking, alcohol consumption, education, sleep difficulty, total sleep time, hypertension, diabetes, and other physical and mental health conditions.The interaction between mild-to-moderate OSA and CBVD was statistically significant ( p = 0.03).The interaction between mild-to-moderate OSA/CBVD and age groups (<60 vs. ≥60 years) was statistically significant (p = 0.05).mmOSA = mild-to-moderate OSA with an AHI 5-29.9 events per hour of sleep.Abbreviations: AHI, apnea-hypopnea index; BMI, body mass index; CBVD, cardiovascular and/or cerebrovascular disease.* Association between mild-to-moderate OSA and comorbid hypertension on all-cause mortality Adjusted for race/ethnicity, sex, BMI, smoking, alcohol consumption, education, sleep difficulty, total sleep time, diabetes, and other physical and mental health conditions.OSA = AHI 5-29.9 events per hour of sleep.