Association of Sedentary Time and Incident Heart Failure Hospitalization in Postmenopausal Women

Introduction

The 2018 revision of the Physical Activity Guidelines for Americans emphasized the importance of reducing time spent in sedentary behaviors (SB), in addition to participating in regular physical activity, for both general and cardiovascular health.¹ At the time the initial 2008 version of the guidelines were written, there was insufficient evidence to support a clear recommendation on the role of physical activity and SB in heart failure (HF) risk.² The 2018 revised guidelines now indicate there is consistent evidence of lower risks of HF development associated with greater amounts of physical activity.³ A recent review of findings from epidemiological cohort studies and pooled analyses published between 2000 and 2017⁴ uncovered only one study on SB and HF incidence, which was in men.⁵ Since that review, one additional study has been published in men and women combined.⁶ No study has focused on older women. The prevalence of HF and its strain on the healthcare system are projected to increase over the next few decades with population aging.⁷ Enhancing current approaches for primary HF prevention is, therefore, paramount.⁸ Given the high prevalence of prolonged sedentary time among US adults aged ≥65,⁹ among whom HF burden is substantial,¹⁰ understanding the role SB has in HF development is relevant to future HF prevention strategies.

We evaluated the prospective association between SB and incident HF hospitalization in postmenopausal women enrolled in the WHIOS (Women’s Health Initiative Observational Study). Our analysis uniquely examined time-varying SB while accounting for several relevant variables including recreational physical activity, measured blood pressure, physical functioning status at baseline, and time-varying coronary events interim to HF hospitalization. Our primary hypothesis was that greater SB would be associated with higher HF risk, independent of the amount of physical activity participation.

Methods

Data that support the findings of this study are available from the corresponding author upon reasonable request and with permission from the WHI.

Study Design and Participants

The WHI design, recruitment, and implementation have been described elsewhere.^11,12 There were 161 808 postmenopausal women, aged 50 to 79 who were without conditions that would limit life expectancy to less than or equal to 3 years, enrolled into either the clinical trials (n=68 132) or observational study (OS, n=93 676) at 40 US centers during 1992 to 1998. The main WHI study ended in 2005. All 161 808 women enrolled in the WHI were followed for incident acute HF hospitalization from study enrollment through September 12, 2005.

The present analysis includes only women in the WHIOS cohort because the SB questionnaire was not administered in the clinical trial. Women with self-reported HF at enrollment (1356) and women reporting the inability to walk one block without assistance (8168) were excluded from these analyses. Further exclusions were made for missing information on SB (934), follow-up after baseline (359), and covariates (1877). The final analytic cohort herein was 80 982 women. Institutional Review Board approval and participant informed consent was obtained at all 40 centers. This study conformed to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for human observational studies. Here, we build upon a previous investigation wherein we demonstrated inverse associations between self-reported physical activity levels and incident HF hospitalization.¹³ Our previous study did not include analysis of SB exposures, which are the focus here.

SB Assessment

SB was assessed by questionnaire at baseline and again at years 3, 6, 9 using 2 questions: (1) “During a usual day and night, about how many hours do you spend sitting? Be sure to include the time you spend sitting at work, sitting at the table eating, driving or riding a car or bus, and sitting up watching TV or talking.” and (2) “During a usual day and night, about how many hours do you spend sleeping or lying down with your feet up? Be sure to include the time you spend sleeping or trying to sleep at night, resting or napping, and lying down watching TV.” There were 8 response categories for each question, ranging from <4 hours to ≥16 hours per day. Total awake SB time was calculated by combining the reported hours of sitting and lying time and then subtracting reported hours sleeping (asked separately from the above questions) as done in previous WHI studies.¹⁴ We also evaluated daily sitting time (hours/day) as a separate exposure. Test-retest reliability for response to the sitting and sleep questions reported 3 months apart in 564 women was Kappa=0.60.¹² This is similar to the test-retest reliability for the SB questions repeated 3 months apart among women in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort (Kappa=0.74).¹⁵ A positive correlation (r=0.31) has been demonstrated between responses to the WHI sitting question and accelerometer measured sedentary time.¹⁶ Both total SB and sitting time were evaluated as time-varying exposures in the present study.

HF Ascertainment

A standardized WHI protocol for identifying incident HF hospitalizations was followed as described in our previous study on physical activity and HF.¹³ Self-reported information on HF hospitalization was collected annually and adjudicated by trained physicians using medical records.¹⁷ Adjudicated cases required a diagnosis of acute HF on the hospital admission, and at least one of the following 4 clinical observations: (1) physician-diagnosis of HF with medical treatment; (2) symptoms plus documentation in the current medical record of a history of imaging procedure showing impaired systolic or diastolic left ventricular function; (3) pulmonary congestion/edema on chest radiograph during current admission; (4) dilated ventricle(s) or poor left or right ventricular function by echocardiography, radionuclide or other contrast ventriculography. This method has good agreement (Kappa=0.79) comparing central and local adjudicated HF¹⁸; and demonstrates modest Kappa coefficients (0.10–0.32), high sensitivity (0.80–0.95), and moderate positive predictive values (0.62–0.68) when compared with other published HF algorithms used in epidemiological studies.¹⁹

Covariate Information

Self-administered questionnaires were completed by participants at baseline to provide information on demographics, history of physician-diagnosis and treatment with medication for diabetes, hypertension, and dyslipidemia, history of atrial fibrillation, stroke, coronary heart disease (CHD; myocardial infarction or coronary revascularization procedures [Coronary Artery Bypass Graft, Percutaneous Coronary Intervention]), hysterectomy status, medication use, and lifestyle behaviors.^11,12 Self-reported total recreational physical activity was summarized in metabolic equivalent hours per week.¹³ Height and weight were measured in clinic and body mass index calculated (kg/m²). Resting blood pressure was measured using a mercury sphygmomanometer and auscultatory methods with cuff size based on measured arm circumference in accord with guidelines contemporaneous to WHI baseline.²⁰ The average of 2 measures was used in analyses. Physical functioning was assessed using the RAND-36 instrument.¹¹

Statistical Analysis

Baseline characteristics were compared according to categories of total SB using a linear (continuous covariates) or logistic (dichotomous covariates) model with the covariate of interest as a function of linear trend across SB categories, using the median value within each. Cox proportional hazards regression was used to estimate hazard ratios (HR) and 95% CIs for incident HF hospitalization in relation to time-varying SB exposures. Ties were handled using the approximate likelihood of Efron.²¹ The proportional hazards assumption was examined by graphically plotting survival by SB categories, and by modeling the HF outcome as a function of the interaction between total SB and follow-up time. In both cases, no appreciable violations were noted. Time-to-event was accrued from date of enrollment in the WHIOS to the date of first hospitalization for acute HF, death, loss to follow-up, or September 12, 2005, whichever came first. Cox regression models were estimated with increasing control for potential confounding, beginning with a model including baseline sociodemographic factors, smoking, alcohol use, menopausal hormone therapy, and hysterectomy status, then adding comorbidities and factors that potentially mediate the association between SB and HF. The primary exposure variable was time-varying total SB time defined categorically as ≤6.5, 6.6–9.5, and >9.5 h/d for consistency with previous WHI studies.²² Time-varying sitting time (≤4.5, 4.6–8.5, and >8.5 h/d) was also evaluated. SB information was updated at years 3, 6, and 9. These updates were incorporated in the time-varying SB models, with the SB variable being updated over the course of follow-up within the model. Tests for linear trend were conducted across median values of categorical SB exposures. Stratification by baseline age, race/ethnicity, body mass index, diabetes, hypertension, CHD, and physical function was used to explore consistency of associations between SB and HF between these specified cohort subgroups. We also stratified on baseline self-reported total recreational physical activity levels to understand whether an association between SB and HF was limited to women who reported a certain amount of physical activity. Tests for interaction were based on the interaction term from a proportional hazards model with HF as a function of SB categories, the subgroup of interest, and their interaction. All baseline covariates remained fixed during follow-up, except CHD, which was a time-varying covariate based on CHD history at enrollment and adjudicated myocardial infarction or coronary revascularization during follow-up interim to HF diagnosis. Sensitivity analyses were conducted to evaluate the potential impact of baseline subclinical morbidity by discarding the first 2 years of follow-up, by excluding women with any difficulty in activities of daily living and those with a self-report of no recreational physical activity, and by stratifying the primary results on baseline physical functioning and CHD history. Population-attributable risk (PAR) percentages were calculated to evaluate the potential impact of risk factor reduction from a public health perspective: PAR%=P_c(1–1/HR_adj)×100%, where P_c is the risk factor prevalence among HF cases and HR_adj is the multivariable-adjusted HR for HF associated with the risk factor.²³P values are for 2-sided hypothesis tests at alpha 0.05, which were conducted using SAS (Windows v.9.4; SAS Institute, Carey, NC).

Results

The cohort at baseline (Table 1) was, on average, 63 years of age and educated beyond high school, with a racial-ethnic distribution of 84.9% White people, 7.1% Black people, 3.4% Hispanic, 2.9% Asian, 0.4% Native American. About 6% and 42% reported current or past smoking, respectively, and 45% reported current menopausal hormone therapy use. Baseline prevalence of diabetes, CHD, and atrial fibrillation was relatively low, whereas more than one-third of women reported treated hypertension. On average, participants reported 7 h/d of total awake SB time and reported high physical functioning. About 11% of the cohort reported no recreational physical activity. All baseline characteristics in Table 1, except for diastolic BP and history of diabetes, CHD, stroke and atrial fibrillation were significantly (P<0.05) associated with total SB, although some of the differences between SB groups were relatively small. Additionally, each baseline characteristic differed significantly according to incident HF hospitalization status (Table I in the Data Supplement).

We identified 1402 (1.7%) cases of incident acute HF hospitalization during 728 838 person-years follow-up. Rates of HF hospitalization (per 1000 person-years) were 1.9, 2.2, and 2.4 over time-varying total SB categories of ≤6.5, 6.6–9.5, and >9.5 h/d, and were 1.9, 2.3, and 2.4 over time-varying sitting categories of ≤4.5, 4.6–8.5, and >8.5 h/d (Table 2). After controlling for demographic factors, smoking, alcohol, menopausal hormone therapy, and hysterectomy, a positive association with incident HF hospitalization was observed with total SB and sitting time (trend, P<0.001 each). Women had a 42% (95% CI, 25%–61%) and 54% (95% CI, 34%–78%) greater HF risk when comparing the highest with lowest category of total SB and sitting time, respectively. The significant associations persisted and were of similar magnitude after controlling further for several comorbid potential mediating factors, including time-varying CHD (model 3; trend, P<0.001 each), and for baseline recreational physical activity (model 4; trend, P<0.001 each).

Stratified associations with incident HF hospitalization are given for total SB in Table 3 and for sitting time in Table II in the Data Supplement. Crude rates of HF hospitalization were positively associated with baseline age, body mass index, treated diabetes and hypertension, and history of CHD and were inversely associated with physical functioning and recreational physical activity. Rates were similar between White and Black women. The multivariable associations with incident HF hospitalization for total SB and sitting time were not significantly different between any of the subgroups evaluated (interaction P>0.05 all).

Sensitivity analyses excluding cases of HF hospitalization occurring in the first 2 years of follow-up and excluding women with functional limitations and those reporting no recreational physical activity were conducted to understand potential reverse causation biases. Multivariable relative risks and their positive trends (P<0.01 all) across categories of SB exposures were consistent with the primary results (Table 4).

For public health context, PAR was estimated to quantify the theoretical proportion of HF hospitalization in the source population that might be averted, assuming a causal association, if women’s baseline total SB was less than the highest exposure category (9.5 h/d). PAR was also estimated for other modifiable HF risk factors obesity, diabetes, hypertension at baseline, and for CHD based on combined information at baseline and interim to HF hospitalization. PARs were 24.1% for hypertension, 15.0% for CHD, 9.6% for obesity, 8.9% for diabetes, and 7.3% for SB (Table 5).

Discussion

The prospective findings from the present study support the hypothesis that greater self-reported SB is associated with significantly higher risk of incident acute HF hospitalization in older postmenopausal women, independent of self-reported physical activity levels. The significant positive associations for time-varying total SB and sitting time with risk of HF hospitalization were evident when controlling for several sociodemographic factors and potential mediating factors measured at baseline and for time-varying CHD events interim to HF hospitalization. When associations were stratified on baseline subgroups, including race/ethnicity, history of treated diabetes, hypertension, and CHD, recreational physical activity, and physical function status, associations were consistent across strata, and tests for interaction were not statistically significant. Despite lack of statistical evidence of an interaction between SB and physical activity, among women in the highest category of physical activity (≥7.5 metabolic equivalent hours per week), which is consistent with guideline recommendations that we previously showed was associated with lower HF risk,¹⁹ those whose total SB exceeded 9.5 h/d had a significant 24% higher multivariable-adjusted risk of incident HF hospitalization. Even greater HF hospitalization risk (42% higher risk) in this physical activity subgroup was evident in women whose sitting time exceeded 8.5 h/d. This suggests that prolonged exposure to SB may substantially elevate risk of developing HF even in older women who report being physically active. The present findings build on and expand those of our previous investigation that demonstrated an inverse association between self-reported physical activity levels and incident HF hospitalization.¹³ As such, efforts to reduce sedentary time and increase activity-related energy expenditure, even through habitual walking,¹³ may be needed for effective HF prevention in later life. Because national prevalence of SB⁹ and physical inactivity²⁴ is high in older women, and because numbers of adults ≥60 are expected to increase from 33 to 79 million between 2010 and 2050,²⁵ improved strategies to reduce SB and increase physical activity in these individuals should enhance public health efforts for HF prevention.

We used PAR, which takes into account both the amount (prevalence) of a risk factor and its relative strength of association (HR) with disease in a defined population, as a theoretical indicator of the potential impact that reducing SB time might have on HF prevention. If women were not exposed to prolonged SB (>9.5 h/d), 7.3% of incident HF hospitalizations might have been avoided. By comparison, PARs for other modifiable HF risk factors were 24.1%, 15.0%, 9.6%, and 8.9% for hypertension, CHD (time-varying), obesity, and diabetes, respectively. Our PAR computation is based on the multivariable-adjusted HR (model 4, Table 2), thus the above PARs control for differences in demographic, clinical, and lifestyle factors, in addition to each of the risk factors for which PAR was computed. The PARs for hospitalized HF estimated in our study on postmenopausal women are consistent with those for hypertension (28%), CHD (16%), obesity (12%), and diabetes (10%) in women in the Olmsted County Study.²⁶ Prolonged SB has relevance to the burden of HF in postmenopausal women enrolled in the WHI similar to that of well-recognized HF risk factors obesity and diabetes. Development of effective approaches for reducing population levels of SB could have considerable impact on HF in later life because SB is also a risk factor for each of the above modifiable HF risk factors.³

We are aware of only 2 published studies focused on the association between SB and HF risk. Young et al⁵ studied over 82 000 men who at enrollment in the California Men’s Health Study were ≥45 years and without known HF. These men self-reported daily time spent watching television, sitting at a computer or reading at baseline, and were then followed a mean of 8 years during which 3472 HF cases were identified. After controlling for several of the same covariates as in our present study, relative risks were 1.09 and 1.27 for the middle (3–4 h/d) and highest (≥5 h/d) compared with the lowest (≤2 h/d; referent) tertile of baseline sitting time (Trend, P<0.001). These associations were of similar magnitude as seen for time-varying total SB and sitting time in the present study (Table 2, model 4). Unlike in our study, where SB was associated with significantly higher HF risk among women reporting the highest amount of recreational physical activity, prolonged sitting was not associated with HF risk among men with high physical activity levels in the California Men’s Health Study. In MESA (Multi-Ethnic Study of Atherosclerosis), spending >1890 minutes/wk (4.5 h/d) in SB was not significantly associated with HF risk when compared with ≤1890 minutes/wk in cohort of 6814 US women and men, who were aged 45 to 84 at baseline and followed 11 years for HF.⁶ This study likely had limited statistical power due to the small numbers of HF (178 cases). Our findings from the WHI and those from the California Men’s Health Study⁵ provide consistent evidence for an adverse association between prolonged SB and HF risk in women and in men.

Evidence suggests that SB is not merely the lower end of the physical activity energy expenditure continuum, but that SB has distinct physiological influences on cardiovascular health.²⁷ The adverse cardiometabolic,²⁸ hemostatic, and inflammatory²⁹ biomarker profiles associated with prolonged SB promote atherosclerosis onset and progression, thus, increasing risk for ischemic HF.³⁰ Mechanisms may also involve pathways influencing myocardial contractility, relaxation, or both. In the landmark 1968 Dallas Bedrest Study, marked reductions in cardiac function were documented in healthy college-aged men following 20 days of voluntary bedrest.³¹ Although bed rest is not completely comparable to habitual SB because of periodic interruption in SB and different gravitational forces, the bedrest study did stimulate interest in the adverse structural and functional cardiac responses to prolonged inactivity. More recently, self-reported SB has been associated with wall stiffening in large conductance arteries,³² reduced arterial distensibility,³³ greater left ventricular mass and mass-to-volume ratio,³⁴ reduced left ventricular compliance,³⁵ and impaired systolic³⁶ and diastolic³⁷ functions. A major contributing factor for vascular dysfunction associated with prolonged SB is reduced arterial wall shear stress secondary to reduced blood flow that results from infrequent skeletal muscle contraction.³⁸ Low skeletal muscle metabolic demand during prolonged SB is accompanied by reduced vasodilatory metabolites, leading to further reductions in blood flow. Reduced arterial blood flow and shear stress lead to uncoupling of endogenous nitric oxide synthase from nitric oxide production and a resulting increase in endothelial cell secretion of endothelin-1, which promotes vasoconstriction. Over time, this can lead to increases in total peripheral resistance, elevated systemic blood pressure, and increased left ventricular afterload. SB is also associated with increased sympathetic nervous system activity.³⁹ Higher sympathetic tone exacerbates vasoconstriction and can lead to reduced renal blood flow and glomerular filtration rate as well as increased renin and aldosterone activity, which increases cardiac afterload and left ventricular myocardial workload. Pericardial ectopic fat deposition is positively associated with time spent sitting⁴⁰ and adversely effects cardiac function.⁴¹

Strengths of the current study include the large cohort size of older postmenopausal women, use of time-varying SB exposures, physician-adjudicated cases of incident acute HF hospitalization, and extensive control for potential confounding factors including time-varying CHD events interim to HF hospitalization. A positive association between time-varying SB and risk of incident HF hospitalization in older women not only builds upon our previous finding of inverse associations between HF risk and physical activity levels¹³ but adds needed evidence on SB and HF to a sparse existing literature. Limitations include assessment of SB using a questionnaire, which inevitably can result in some degree of exposure misclassification. Because of the prospective study design, wherein SB assessment preceded disease occurrence, any misclassification would most likely be nondifferential resulting in associations biased toward the null.²³ Baseline prevalence of diabetes, CHD, and atrial fibrillation were lower than what might be expected for women in the age range studied. Women who chose to enroll in the WHIOS might have been somewhat healthier than the general population of women of similar age, which could have resulted in a lower rate of incident HF hospitalization, and possibly weaker associations with SB. Recent studies using accelerometer measures of SB, including within WHI,⁴² suggest that the pattern of SB accrual (eg, bout frequency and duration) as opposed to only total SB time is important to cardiovascular health. Questionnaires are not able to discern SB patterns, and therefore we could not evaluate their relevance to HF risk in the present study. Reverse causation bias is a reasonable concern when prospectively assessing HF risk in relation to SB, which might be influenced by reductions in energy levels and aerobic capacity during the preclinical phases of HF.⁴³ We cannot completely rule this out as an alternative explanation for the positive associations we observed, though associations persisted after excluding HF cases that occurred during the first two years of follow-up and did not differ significantly across physical functioning categories. Our previous study showed recreational physical activity was inversely associated with HF with reduced and preserved ejection fraction¹³ in a subcohort of WHI women for whom records on echocardiographic and other imaging information was available to allow HF subtype determination.³⁰ Unfortunately, we did not have sufficient information to evaluate SB and HF subtypes here.

In conclusion, prolonged SB is associated with a significantly increased risk of incident HF hospitalization in older postmenopausal women. These findings provide evidence for an additional adverse cardiovascular outcome of prolonged SB that could help refine future physical activity guidelines. Results of this study underscore the need for effective strategies to reduce daily SB time, in addition to increasing recreational physical activity, as part of population efforts for HF prevention.

Supplemental Material

List of WHI Investigators.

Tables I and II.

Footnotes