Cardiovascular Autonomic Nervous System Function and Hip Fracture Risk

Abstract

Purpose:

Autonomic nervous system function modulates bone remodeling in rodent osteoporosis models. We tested whether cardiovascular autonomic function is associated with hip fracture risk in humans.

Methods:

Participants were 1299 subjects from the Cardiovascular Health Study (mean age 72.8 years). Eight heart rate variability (HRV) measures (time and frequency domains, detrended fluctuation analysis variables, and heart rate turbulence) were derived from 24-hour Holter monitor scans in sinus rhythm. Median follow up for incident hip fracture was 14.7 years [IQR 9.1, 20.2]. Cox proportional hazards models were used to calculate hazard ratios (95% Confidence Intervals, CI).

Results:

There were 144 hip fractures among 714 women (1.31 [1.06, 1.61] per 100-person years) and 46 among 585 men (0.62 [0.43, 0.90] per 100 person-years). From among HRV variables examined, a one standard deviation (SD) higher variation between normal heart beats over 24 hours (the SD of NN intervals [SDNN]) was associated with a multivariable-adjusted lower hip fracture risk (HR=0.80; 95% CI 0.65-0.99; p=0.04) in women. The adjusted association between very low frequency power and hip fracture was borderline statistically significant in women (HR=0.82; 95% CI, 0.66-1.00; p=0.06). When the 8 HRV variables were considered conjointly and adjusted for each other’s association with hip fracture risk, a 1 SD higher SDNN value was significantly associated with reduced hip fracture risk in women (HR 0.74; 95% CI, 0.50-0.99; p=0.05). No HRV variables were associated with hip fracture in men.

Conclusions:

In older women, increased heart rate variation is associated with hip fracture risk.

PRECIS

Among 1299 older adults with 24-hour Holter monitoring data at baseline, followed for approximately 15 years, 190 incident hip fractures occurred. Increased heart rate variability was independently associated with reduced risk of hip fracture among female participants.

INTRODUCTION

Several lines of evidence suggest the autonomic nervous system (ANS) may play a role in bone remodeling and maintenance of bone mass, although much of the evidence comes from rodent experiments. The effects of the ANS on bone may be direct, through nerve fibers reaching bone cells or through diffusion of neurotransmitters activating adrenergic receptors on cells involved in bone physiology (1). Alternatively, the effect may be indirect, such as through increased sympathetic tone. As examples, the sympathetic branch (SNS) of the ANS, through β2-adrenergic receptors, can promote osteoclast differentiation, inhibiting osteoblast proliferation (2). The parasympathetic branch (PNS) of the ANS upregulates osteoclast apoptosis, favoring bone mass accretion (3, 4). Experimentally, severing the sub-diaphragmatic vagus nerve in rodents results in reduced lumbar vertebra bone mass, suggesting the PNS has a positive effect on bone mineral density (5). The interaction of the SNS and the PNS branches of the ANS in the regulation of bone homeostasis, however, is complex and many aspects of its control of bone remodeling are not yet known, especially in people. In humans, for example, enhanced SNS tone gives rise to hypertension (HTN), which is a risk factor for loss of bone density (6).

Heart rate variability (HRV), derived from 24-hour ECG recordings, is a measure of cardiovascular ANS functioning reflecting patterns of autonomic signals (7). HRV has been associated with important cardiovascular outcomes, but its association with bone-related clinical outcomes, like hip fracture, is uncertain. In the present exploratory study, we examine the hypothesis that HRV is independently associated with hip fracture risk in the Cardiovascular Health Study (CHS), a prospective observational study of older adults.

METHODS

The CHS is a longitudinal study of community dwelling adults aged ≥65 years from four US communities drawn from Medicare lists (8). In 1989-1990, 5201 participants were recruited, of whom 246 were African American. To increase African American representation, an additional 687 mostly African Americans were recruited in 1992–1993. All participants gave informed consent upon study entry. Institutional review board approval was received at all clinical sites. From 1989-1990 to 1998-1999, participants were seen in clinic annually and had telephone contact midway between clinic visits. Following the 1998-1999 visit, participants continued to be contacted biennially to update hospitalizations, incident diagnoses, and medications. Surveillance for these analyses ended June 30, 2015.

Analytic Cohort:

Eligible participants for these analyses were CHS participants from 1989-1990 and who completed 24-hour Holter ECG monitoring at that visit (hence the mostly African American cohort, recruited in 1992-1993, was not included in these analyses). Recordings needed to be of good quality and in predominantly normal sinus rhythm, to permit research-quality analysis of HRV values. Twenty-four-hour Holter recordings were obtained at the time of study entry in a subset of volunteers (n=1,429) from all four clinical sites. These volunteers were similar to the total cohort other than that they were younger (71.9 vs 73.2 years of age) and more likely to be female (46.6% vs 41.7%) (9). Of these, 1299 had full data on all covariates and serve as the analytic cohort.

Hip Fractures:

Following the 1998-1999 visit, fracture data were obtained from patient-report and then confirmed through hospital medical records, including discharge summaries, gathered every 6 months through June 30, 2015. Data were checked against Medicare claims data to identify any hospitalizations not reported by the participant. Hip fracture was defined using the International Classification of Diseases, Ninth Revision (ICD-9) code 820.xx. Pathological fractures (ICD-9 code 773.1x) and motor vehicle accidents (E810.xx-E825.xx) were excluded.

Holter Monitoring:

Holter tapes were recorded on Del Mar Avionics recorders, which have a calibrated timing signal. Data were processed by research technicians at the Washington University School of Medicine Heart Rate Variability Laboratory (St. Louis, Missouri), using a GE Marquette MARS 8000 Holter Analyzer (GE-Marquette, Milwaukee, Wisconsin). The longest and shortest true normal-to-normal (NN) heartbeat intervals and the highest and lowest ratio between adjacent NN intervals were identified for each tape, and intervals outside of these limits, including blocked atrial premature beats, were excluded from HRV calculations (10).

HRV Variables:

Heart rate variability (HRV) characterizes changes in the time intervals between consecutive heartbeats. The oscillations of a healthy heart are constantly changing, which allow the cardiovascular system to rapidly adjust to physical and psychological challenges to homeostasis. The following HRV variables were calculated (11):

A. Time domain variables are calculations based on the intervals between normal heart beats (measured in milliseconds [ms]). In the case of abnormal cardiovascular autonomic function, time domain HRV is decreased (i.e., less heart rate variation) and mean heart rates are higher.

AvgNN – average time between NN (normal to normal) beats over 24 hours, in ms. It can be converted to average heart rate (60000/AvgNN) in beats/minute.

SDNN – the standard deviation of all NN intervals over 24 hours. It is a measure of overall heart rate variability. Better health (e.g., good quality sleep, more physical activity) is associated with greater SDNN variability.

B. Frequency domain variables are derived by deconstructing the total variance in heart rate patterns into their underlying oscillatory components. For this study we examined low frequency (LF) power [0.04 to 0.15 Hz (3-9 cycles per minute)] and very low frequency (VLF) power [0.0033 to 0.04 Hz (one cycle every 20 seconds to every five minutes)]. ANS dysfunction is associated with a reduction in the proportion of time an individual spends with these heart rate patterns reduction. Frequency domain HRV is measured in milliseconds squared (ms²) and is generally ln transformed to permit parametric statistical calculations.

Low frequency (LF) power is related to both SNS and PNS activity. LF power primarily reflects baroreceptor activity while at rest. Baroreceptors are stretch-sensitive mechanoreceptors located in the chambers of the heart and vena cavae, carotid sinuses and the aortic arch. Decreased baroreflex function is related to aging and to impaired regulatory capacity.

Very low frequency (VLF) power reflects mechanisms to thermoregulation, the renin-angiotensin system and hormonal factors. Low mean power in this band has been associated with inflammation (12) and has been correlated with low levels of testosterone (13). Changes in heart rate in this band are also associated with sleep disordered breathing and periodic limb movements. Administration of atropine blocks VLF power (14); VLF is therefore considered a measure of PNS activity.

C. Non-linear HRV (15) is a family of measures that reflects the structure of HR patterns at different scales. In the current analysis, a measure called detrended fluctuation analysis (DFA) was tested. It reflects the randomness vs. the self-affinity of the heart rate patterns within of each sequence of beats compared with the subsequent beats. DFA1 is on a scale of 4-11 beats; DFA2 is on a scale of 12-20 beats. DFA values range from 0.5, i.e., completely random, to 1.5, i.e., completely predictable. Normal values of 1.0 means heart rate variability of 50% random signals and 50% recurring signals. Values above 1.0 mean more stability and indicate compensation processes within the autonomic control systems. Values below 1.0 suggest greater randomness in heart rate patterns. Values are correlations, so they are unitless.

D. Heart Rate Turbulence (HRT) (16) is a measure of the resilience of the ANS. It requires at least 5 isolated premature ventricular beats (PVBs) on a Holter recording for calculation and measures how well the ANS responds to the loss of cardiac output that occurs after a PVB. Results are averaged over all the PVBs. Participants without at least 5 isolated PVBs on a Holter recording are considered to have normal HRT. Typically, heart rate speeds up after a PVB, to compensate for the loss of cardiac output during the PVB, and then slows down, returning to the original heart rate. The two phases of HRT, rate acceleration and deceleration, are quantified by 2 parameters termed turbulence onset (TO) and turbulence slope (TS).

1. TO (turbulence onset) measures the acceleration in HR as an immediate response to the loss of cardiac output after a PVB and is a percent change of the NN interval 2 beats after the PVB compared to the two beats before the PVB. In healthy hearts, which accelerate after a PVB, shortening the NN interval, this index is negative. If the TO >=0 it is abnormal because the NN interval did not go down, i.e., there was no compensatory increase in heart rate after the loss of cardiac output with the PVB (15).

2. TS (turbulence llope) is the slope of the steepest regression line fitted over 5 consecutive sinus rhythm NN intervals within the 15 NN intervals after the PVB. It is a measure of the deceleration phase, during which the NN interval becomes longer (more ms between each beat).TS is measured in ms/beat. A TS value <3 was considered abnormal in CHS since it maximally separates participants who do and do not die of CVD in CHS (17).

Covariates:

Multivariate analyses were adjusted for factors that could impact HRV and/or osteoporosis. These included sex, age, race and clinic site; history of hypertension, hypertension medications (including beta blockers and thiazides), diabetes, smoking status (current, never, former), current alcohol use (never, ≤7 drinks per week, >7 drinks per week), weight, height, difficulties with activities of daily living (ADL) and instrumental ADL (IADL), lipid lowering medication use, bisphosphonate use, prevalent coronary heart disease (CHD; prior myocardial infarction, angina, CABG, or angioplasty), estimated glomerular filtration rate based on cystatin (eGFRcyst), and log transformed high-sensitivity C-reactive protein levels (CRP). A physical activity (PA) score was calculated based on leisure-time activity (ordinal score of 1-5 for quintiles) and pace of walking (ordinal score of 1-3 for pace <2 mph, 2-3 mph, or >3 mph) into a single physical activity score (18).

At baseline, participants were also queried regarding sleeping habits. Questions included whether one felt groggy or sleepy upon awaking in the morning; whether a spouse or roommate complained of the participant snoring or stopping to breath during sleep; trouble falling asleep; and waking up several times at night.

In 1992/1993 (up to 4 years after Holter monitoring was performed) osteocalcin (OC) and C terminal cross-linking telopeptide of type 1 collagen (CTX) were measured from fasting serum specimens in an ancillary study of CHS women (n=1680) (19). OC is an osteoblast marker of bone formation (20); CTX is a measure of bone resorption (21). There were 442 women in the present study who had these analytes measured.

Statistical Analysis:

We compared the characteristics of CHS participants with Holter monitoring and full data on covariates (n=1299) versus those without Holter monitoring (n=3772) or those with Holter monitoring but without full data on covariates (n=130). t tests for continuous variables and chi-square tests for categorical variables were used. Time to incident hip fracture was calculated as the interval from the 1989/1990 study visit to the earliest date of first incident hip fracture, death, loss to follow-up, or end of follow-up on 6/30/2015. Cox proportional hazards models per 1 SD were generated for each HRV variable to estimate the hazard ratio (HR) of incident fracture. The HRV variables were also examined together in a single model for risk of incident hip fracture adjusting for the relationship of each variable with fracture risk.

Analyses were performed separately for men and women given the differences in the pathophysiology and incidence rates of hip fractures by sex. Statistical significance was set at p<0.05.

We used nested models adjusted for factors as: M0: unadjusted; M1: age, race and clinic site; M2: M1+ hypertension, hypertension medications, diabetes, smoking status (current, never, former), current alcohol use (never, ≤7 drinks per week, >7 drinks per week), weight, height, difficulties with ADL and IADL, lipid lowering medication use, prevalent CHD, eGFRcyst, and CRP (log transformed).

Analyses were performed using the R statistical package 2019 (22).

RESULTS

Baseline characteristics of CHS participants who underwent Holter monitoring who were in predominately normal sinus rhythm and who had complete covariates compared to other participants are shown in Supplemental Table 1. These participants (n=1299) were healthier than those who did not undergo Holter monitoring (n=3772) or who had Holter monitoring but incomplete data on covariates (n=130). They were younger; had lower systolic blood pressure; had better renal function, less difficulty with ADL and IADL; were less likely to be current smokers, had higher body-mass index, and were less likely to report fair or poor health.

Characteristics of the analytic cohort categorized by sex are shown in Table 1. Compared with women, men were older; had higher diastolic blood pressure; had more education; and had more diabetes, past smoking, and CHD; worse renal function; less use of bisphosphonates; less use of thiazides and beta blockers; but more physical activity. Women had more difficulties with IADL though they did not differ from men regarding ADL or self-reported perceived health. Regarding sleep, men snored more often and reported more trouble sleeping than women. There were no differences regarding the number of self-reported falls in the year prior to the baseline examination.

Table 1:

Baseline characteristics of the analytic cohort categorized by sex.

	Women N = 714	Men N = 585	P Value
Demographic & Medical Conditions
Age (years)	71.3 ± 4.7	72.3 ± 5.0	<0.001
BMI (kg/m2)	26.7 ± 4.7	26.6 ± 3.5	0.67
SBP (mmHg)	134.2 ± 20.8	134.8 ± 2.8	0.59
DBP (mmHg)	68.9 ± 10.6	71.8 ± 10.8	<0.001
Education ≥12 years (%)	43.1	50.3	0.01
Diabetes (%)	11.8	18.8	0.001
African American (%)	4.3	4.3
Smoking Status (%)			<0.001
Current	10.6	7.7
Past	31.4	60.8
Never	58.0	31.5
Current Alcohol Use (%)			0.98
<7 drinks per week	89.2	89.4
≥7 drinks per week	10.8	10.6
Medication Use (%)
HTN Medications	44.3	43.6	0.84
Lipid Medications	5.0	5.0	1.00
Estrogen (females)	15.8	--	<0.001
Beta blocker	12.6	16.8	0.052
Bisphosphonate	4.0	0.3	<0.001
Thiazide	22.4	15.2	0.002
Laboratory
eGFR based on cystatin-C levels (ml/min/1.73m2)	81.9 ± 18.9	74.8 ± 17.3	<0.001
Log2 (CRP) mg/L	1.3 ± 1.5	1.2 ± 1.5	0.13
Prevalent CVD (%)
CHD	13.7	26.5	<0.001
Stroke	2.7	4.8	0.06
TIA	1.8	3.6	0.07
QOL
Difficulty with ≥ 1 ADL (%)	5.2	5.2	1.00
Difficulty with ≥1 IADL (%)	25.7	14.1	<0.001
Self-reported health (%)			0.82
Excellent, very good, good	81.2	80.5
Fair, poor	18.8	19.5
Physical Activity (%)			<0.001
Low	24.8	16.0
Medium	56.6	51.2
High	18.6	32.8
Sleeping Issues
Groggy in daytime	10.9	7.2	0.03
Trouble breathing in sleep - observed by bedmate	3.7	14.6	<0.001
Wake up several times at night	62.8	64.5	0.46
Snoring	20.5	34.6	<0.001
Self -Reported Falls (%)
More than one fall in year prior to baseline (%)	3.0	2.3	0.52

ADL – activities of daily living; BMI – body mass index; CHD – coronary heart disease; CHS – Cardiovascular Health Study; CRP – C reactive protein; CVD – cardiovascular disease; DBP – diastolic blood pressure; eGFR – estimated glomerular filtration rate; HTN – hypertension; IL-6 – interleukin 6; IADL – instrumental activities of daily living; QOL – quality of life; SBP – systolic blood pressure; TIA – transient ischemic attack

Mean and median HRV values appear in Supplemental Table 2. Most mean HRV values were significantly lower for women than men. Correlation coefficients between HRV variables are shown in Supplemental Figure 1. SDNN was strongly correlated with DFA1 and VLF.

The mean follow-up time was 14.7 years (SD 6.9) [median 14.7 years [interquartile range, 9.1, 20.2]). The mean time to fracture was 14.2 years (SD 7.0) (median 14.1 years [interquartile range, 8.2, 19.6]). There were 190 incident hip fractures among the 1299 participants in the cohort: 144 among 714 women (incidence rate 1.31 [1.06, 1.61] fractures per 100 person years) and 46 among the 585 men (0.62 [0.43, 0.90] fractures per 100 person years).

Hazard ratios (HR) for each HRV parameter for every 1 SD higher value are shown in Table 2. In fully adjusted models, SDNN was associated with a significantly reduced risk of hip fracture (HR 0.80 [0.65, 0.98]; p=0.03) and VLF had a borderline lower risk of hip fracture (HR 0.82 [0.67, 100]; p=0.06) among women. Among men there was a significantly lower hip fracture risk for a 1 SD increase of VLF in non-fully adjusted models. After full adjustment for covariates, the HR for hip fracture was essentially unchanged but the confidence intervals were no longer statistically significant. A 1 standard deviation higher heart rate (NN) was not associated with fracture risk.

Table 2:

Hazard ratios for each HRV variable considered separately for incident hip fracture risk. Values are per one standard deviation increase in the value of the HRV variable. LF and VLF are Ln transformed.

	Female				Male
HRV variable	HR	95% CI	P value		HR	95% CI	P value
M0
NN	1.00	0.83, 1.21	0.98		0.96	0.73, 1.26	0.75
SDNN	0.79	0.66, 0.96	0.02		1.07	0.81, 1.41	0.63
LF	0.87	0.72, 1.04	0.14		0.84	0.62, 1.16	0.29
VLF	0.80	0.66, 0.96	0.02		0.72	0.52, 1.00	0.05
DFA1	0.87	0.72, 1.05	0.15		0.78	0.58, 1.06	0.11
DFA2	0.99	0.82, 1.21	0.98		0.91	0.65, 1.26	0.57
TO	0.96	0.79, 1.16	0.65		1.07	0.78, 1.47	0.70
TS	0.93	0.76, 1.14	0.47		0.93	0.68, 1.28	0.66
M1
NN	0.96	0.79, 1.17	0.69		0.90	0.67, 1.19	0.45
SDNN	0.81	0.66, 0.98	0.03		1.07	0.81, 1.14	0.63
LF	0.91	0.76, 1.10	0.33		0.97	0.69, 1.35	0.86
VLF	0.82	0.68 1.00	0.05		0.71	0.51, 1.00	0.05
DFA1	0.90	0.75, 1.09	0.29		0.86	0.63, 1.18	0.35
DFA2	0.97	0.81, 1.17	0.75		0.93	0.67, 1.28	0.64
TO	0.93	0.77, 1.14	0.50		1.06	0.76, 1.48	0.73
TS	0.97	0.80, 1,17	0.73		0.94	0.58, 1.31	0.72
M2
NN	0.95	0.77, 1.17	0.64		1.02	0.71, 1.45	0.93
SDNN	0.80	0.65, 0.99	0.04		1.32	0.94, 1.85	0.11
LF	0.89	0.76, 1.08	0.24		0.75	0.48, 1.15	0.19
VLF	0.82	0.66, 1.00	0.06		0.89	0.58, 1.37	0.59
DFA1	0.87	0.71, 1.07	0.19		0.72	0.46, 1.12	0.15
DFA2	0.95	0.79, 1.15	0.91		0.84	0.52, 1.33	0.46
TO	0.94	0.76, 1.15	0.52		0.91	0.60, 1.38	0.65
TS	0.94	0.77, 1.14	0.50		1.15	0.79, 1.67	0.46

M0 – unadjusted model; M1 – model adjusted for sex, age, race and clinic site; M2: model M1 further adjusted for hypertension, hypertension medications, diabetes, smoking status (current, never, former), current alcohol use (never, 7 or fewer drinks per week, >7 drinks per week), weight, height, difficulties with ADL, lipid medication use, prevalent CHD, eGFRcyst, and CRP log transformed.

DFA - detrended fluctuation analysis; LF – low frequency domain; NN - average time between NN (normal to normal) beats over 24 hours; SDNN - the standard deviation of all NN intervals over 24 hours; TO – turbulence onset; TS – turbulence; VLF – very low frequency domain

Conjoint Model (Table 3):

Table 3:

Hazard ratios for HRV variables with incident hip fracture risk when all HRV variables were considered together in one model. Values are per one standard deviation increase in the value of the HRV variable. LF and VLF are Ln transformed.

	Female				Male
HRV variable	HR	95% CI	P Value		HR	95% CI	P Value
M0
NN	1.21	0.87, 1.68	0.25		0.98	0.58, 1.67	0.94
SDNN	0.72	0.52, 1.00	0.05		1.35	0.88, 2.07	0.17
LF	1.30	060, 2.84	0.51		1.74	0.44, 6.86	0.43
VLF	0.94	0.65, 1.37	0.76		0.63	0.36, 1.12	0.11
DFA1	0.76	0.32, 1.83	0.54		0.51	0.11, 2.39	0.39
DFA2	0.84	0.59, 1.20	0.35		1.24	0.60, 2.56	0.57
TO	0.92	0.75, 1.14	0.43		1.09	0.75, 1.56	0.66
TS	0.91	0.72, 1.15	0.41		1.11	0.73, 1.58	0.58
M1
NN	1.09	0.78, 1.51	0.63		0.90	0.53, 1.53	0.70
SDNN	0.73	0.53, 1.01	0.06		1.43	0.91, 2.23	0.12
LF	1.24	0.57, 2.72	0.59		1.84	0.49, 6.90	0.37
VLF	0.99	0.69, 1.44	0.99		0.60	0.34, 1.06	0.08
DFA1	0.85	0.35, 2.05	0.71		0.57	0.13. 2.61	0.47
DFA2	0.81	0.58, 1.13	0.21		1.09	0.53, 2.24	0.81
TO	0.92	0.75, 1.14	0.45		1.15	0.80, 1.66	0.46
TS	0.95	0.76, 1.17	0.62		1.10	0.77, 1.58	0.59
M2
NN	1.04	0.73, 1.48	0.76		0.70	0.35, 1.41	0.32
SDNN	0.74	0.50, 0.99	0.05		1.85	0.96, 3.22	0.06
LF	1.30	0.58, 2.95	0.53		0.96	0.15, 6.13	0.39
VLF	1.09	0.75, 1.59	0.66		0.62	0.30, 1.27	0.19
DFA1	0.70	0.28, 1.77	0.57		0.67	0.07, 6.12	0.72
DFA2	0.85	0.60, 1.19	0.35		1.66	0.57, 4.80	0.35
TO	0.94	0.76, 1.16	0.50		1.15	0.73, 1.80	0.55
TS	0.92	0.73, 1.15	0.44		1.44	0.91, 2.28	0.12

M0 – unadjusted; M1 – adjusted for sex, age, race and clinic site; M2: further adjusted for hypertension, systolic and diastolic blood pressure, hypertension medications, diabetes, smoking status (current, never, former), current alcohol use (never, 7 or fewer drinks per week, >7 drinks per week), weight, height, difficulties with ADL, lipid medication use, prevalent CHD, eGFRcyst, and CRP log transformed.

DFA - detrended fluctuation analysis; LF – low frequency domain; NN - average time between NN (normal to normal) beats over 24 hours; SDNN - the standard deviation of all NN intervals over 24 hours; TO – turbulence onset; TS – turbulence; VLF – very low frequency domain

When the eight HRV variables were considered together, adjusted for each other’s association with hip fracture risk and for covariates, a 1 SD higher SDNN level was significantly associated with lower hip fracture risk among women but not in men. None of the other HRV variables, including NN, was related to hip fracture risk.

Kaplan Meier Plots:

Hip fractures by sex specific SDNN quartiles are shown in Figure 1. Among women, the log rank test was p=0.1; for men p=0.7. Incidence summaries of hip fractures by quartiles of SDNN are shown in Supplemental Table 3. There were more fractures among women in the lower two quartiles of SDNN than in the upper two quartiles of SDNN. Among men, there was an opposite trend with more fractures in the highest SDNN quartile.

FIGURE:

Kaplan Meier plots of hip fractures by sex specific SDNN quartiles. Log rank test was p=0.1 for women; p=0.7 for men.

Factors Associated with SDNN Quartiles (Table 4):

Table 4:

Baseline characteristics of the analysis cohort categorized by sex and quartiles of SDNN variation.

	MEN					WOMEN
	Q1 N=144	Q2 N=143	Q3 N=143	Q4 N=143		Q1 N=176	Q2 N=175	Q3 N=175	Q4 N=175
Demographics
Age (years)	72.7	72.8	71.6	72.2		71.6	71.7	70.9	70.7
Weight (kg)	81.4	80.0	79.6	77.6 **		70.4 ©	67.9	67.0	67.0
SBP (mmHg)	137.7	132.9	134.1	135.0		135.3	133.6	135.0	132.0
DBP (mmHg)	72.8	71.4	72.8	70.7
Current smoker (%)	13.2 †	5.6	7.0	4.9		18.8 ***	8.6	7.4	8.0
Difficulty ≥1 ADL (%)	4.9	5.6	2.8	7.0		8.5	4.6	2.3	5.7
Difficulty ≥1 IADL (%)	16.9	16.8	9.2	14.0		31.2	26.3	22.9	22.4
Reported good health (%)	79.0	79.7	84.5	79.0		77.1	82.9	83.9	81.7
African American (%)	6.2	3.5	1.4	1.6		7.4	4.6	3.4	1.7
Laboratory
Log2 (CRP) (mg/L)	1.8 *	1.1	1.1	0.8		1.6	1.4	1.4	0.9 *
IL6 (pg/ml)	2.78 ††	1.95	2.17	2.00
						N=114	N=103	N=107	N=118
OC (ng/ml)	--	--	--	--		25.3	24.0	23.8	24.3
CTX (ng/ml)	--	--	--	--		0.4	0.4	0.4	0.4
Medical History (%)
Prevalent CHD	32.6	28.7	23.1	21.7		19.3	13.7	10.3	11.4
Stroke	6.2	4.2	4.2	4.2		4.0	1.1	2.9	2.3
TIA	4.9	5.6	3.5	0.7		3.4	1.1	0.6	1.7
HTN	60.8	53.1	54.5	46.2		60.2	52.6	58.6	49.7
Renal Function (ml/min/1.73m2)
eGFRcys	69.5 ©	74.4	75.6	78.6		81.5	81.5	82.5	81.8
Medication Use (%)
Thiazide use	17.4	14.0	15.4	14.0		23.9	19.4	25.7	20.6
Beta blocker use	20.8	16.8	16.1	13.3		20.5	11.4	12.6	5.7 *
Estrogen use (female)	--	--	--	--		21.0	12.6	16.0	14.4
Bisphosphonate use	0.0	0.0	0.0	1.4		5.1	2.9	4.6	3.4
Sleeping Issues (%)
Groggy in daytime	10.6	5.8	5.0	7.1		12.1	10.3	10.1	11.1
Trouble breathing in sleep - observed by bedmate	27.6	26.9	27.2	27.7		4.0	4.1	5.3	1.3
Wake up several times at night	65.3	58.6	69.8	57.6		65.3	58.6	69.8	57.6
Snoring	37.5	37.2	32.8	30.7		20.9	23.8	26.6	10.8 ©©
Physical Activity Score (%)
Low	27.3 *	14.0	9.9	12.7		25.3	26.0	26.6	21.4
Medium	49.0	54.5	57.0	44.4		60.0	57.2	53.2	56.1
High	23.8	31.5	33.1	43.0		14.7	16.8	20.2	22.5
>1 Fall within One Year of Baseline (%)
	3.5	1.4	2.8	1.4		3.5	1.7	2.9	4.0

^*P<0.001

^**p=0.007

^***p=0.001

^†p=0.03

^††p=0.005

^©p=0.01

^©©p=0.003

OC – osteocalcin CTX - C terminal cross-linking telopeptide of type 1 collagen

To understand how higher SDNN variation was associated with lower hip fracture risk we examined baseline characteristics of the cohort by quartile of SDNN. Among women, higher SDNN was associated with slightly younger age, lower weight, and lower DBP; lower CRP and IL6 levels, and less diabetes and less beta blocker use compared to other SDNN quartiles. Higher SDNN levels were associated with less estrogen use as compared to lower SDNN levels. Prevalence across SDNN quartiles was similar for of cardiovascular disease, HTN, self-reported health, falls in the last year, renal function, thiazide and bisphosphonate use, and OC/CTX levels.

Largely similar results were found among men, though men with high SDNN had slightly higher eGFR values. There were no differences regarding use of thiazide diuretics or beta blockers by SDNN quartile.

In men or women, there were no differences in ADL or IADL, trouble sleeping, or trouble breathing during sleep by SDNN quartiles. There were no differences in degree of physical activity among women by quartile of SDNN, though men in the highest SDNN quartile tended to exercise more vigorously than other SDNN strata. Low SDNN was associated with current smoking in men and women.

DISCUSSION

In this exploratory analysis, a one SD higher 24-hour SDNN value in women was significantly associated with lower risk of hip fractures. The beta coefficients of the SDNN with hip fracture risk were essentially unchanged by covariate adjustments, suggesting that the observed associations were not attributable to measured confounders. When all eight HRV factors were considered together and adjusted for each other’s association with hip fracture risk, higher SDNN variation was independently associated with reduced hip fracture risk in women and had a borderline increased association in men. Taken together, these results suggest that increased HRV may impact hip fracture risk, at least in women. To our knowledge, this is the first time that HRV measures have been related prospectively to hip fracture risk. Two prior small studies have demonstrated decreased bone mineral density in association with autonomic dysfunction (23, 24).

SDNN reflects many of the cyclic and diurnal components that make up HRV. The main inputs are the SNS and PNS, baroreflexes, thermoregulation, hormones, sleep-wake cycle, meals, physical activity, and stress. Healthy individuals who exercise (and have low HRV during exercise) and who sleep deeply without interruption (with low heart rate and high HRV) have a wide distribution of HRV. Stressed individuals, those with obstructive sleep apnea, and / or chronic pain have disturbed sleep and elevated heart rates and reduced HRV. It may then be argued that SDNN is a marker of good health. While this may be true, analyses accounted for difficulties with ADL and IADL, prevalent CHD and eGFRcyst, and there were no differences by SDNN quartile for self-reported health, exercise, or difficulty sleeping or waking. It is unlikely then that factors reflecting good health strongly confounded the association of SDNN with fracture risk in women.

An explanation for the association of increased SDNN with reduced fracture risk in women can be posited, although it is necessarily speculative. SDNN and other measures of HRV decrease with age (25). Hence, higher SDNN values in older adults suggest less “biological” aging (26). In younger people, the main regulator of the cardiovascular system is the PNS. In contrast, SNS tone dominates in older people (27, 28). Lower PNS activity, not increased SNS tone, appears to account for reduced HRV with aging (27). The transition from PNS to SNS autonomic control of cardiovascular function may contribute to the increased presence of hypertension with age (27, 28), which in turn is associated with reduced bone mineral density and fracture risk (29, 30). Among women in the highest quartile of SDNN the prevalence of HTN was lower as compared to other quartiles. Further studies will be needed to investigate the notion that elevated SDNN levels are associated with less biological aging.

SDNN values were not associated with hip fracture risk in men on univariate analysis but had a borderline association with increased hip fracture risk when the association of all HRV variables with hip fracture risk was examined. This finding may be due to chance (the association of SDNN with hip fracture risk was not significant on univariate analysis), but there were more fractures in men in the highest quartile of SDNN than in other quartiles. The reason for this discrepancy between men and women is not clear but men and women differed in HRV measures (Supplemental Table 2) (31) and risk factors for osteoporosis and fracture differ by sex (32).

The Study of Osteoporotic Fractures (33) reported that osteoporotic fracture risk increases with elevated heart rate. It reported that women with resting heart rates ≥80 beats per minute had an adjusted 1.6-fold (95% CI, 1.2-2.0) increased risk of a hip, pelvis, or rib fracture and a 1.9-fold (95% CI, 1.4-2.5) risk of vertebral fracture. There was a linear relationship with hip fracture risk by quartiles of HR (60-69, 70-79, 80-89, ≥90). Such an effect was not detected in this study nor has been reported elsewhere. It requires further evaluation.

Several clinical observations support the hypothesis that ANS signals contribute to the regulation of bone remodeling. Low levels of the PNS cholinergic agonist, nicotine, activate the nicotinic receptors, upregulating osteoblast proliferation; high levels of nicotine, as in cigarette smokers, desensitize nicotinic receptors, thereby downregulating osteoblasts (34). People who smoke have more osteoporosis than non-smokers (35). In people with pheochromocytoma, increased catecholamine levels are associated with an increase in bone resorption markers, which are normalized by adrenalectomy (36). Last, a study reported increased sympathetic activity and reduced parasympathetic activity in postmenopausal women with osteoporosis compared with postmenopausal women without osteoporosis (23).

This study has several strengths aside from its long follow up. HRV variables were calculated to research standards and were based on 24-hour Holter scans. This allowed us to gauge the association of HRV variables with hip fracture risk using a broad array of HRV variables. The number of hip fractures was relatively large in women, highlighting its high incidence among older women. We adjusted for many variables that were associated with autonomic dysfunction and hip fracture risk. This study also has limitations. We examined HRV variables only once. HRV values change over time (37). There are no data to our knowledge that examine the impact of change in HRV values on clinical outcomes over long periods of follow up. Second, analyses did not account for change in age or other covariates during the long follow up period. Third, only a limited number of OC and CTX values among women were available. Hence, we could not explore with certainty whether the association of HRV with fracture risk was through bone remodeling. Fourth, we adjusted for multiple factors associated with autonomic dysfunction and bone disease, but residual confounding by unmeasured factors is still possible. Also, some factors (e.g., CRP (38)) may have been mediators between autonomic dysfunction and fracture risk, although the observed association of SDNN with lower risk among women remained even with adjustment for CRP. Fifth, we had no data on concomitant bone mineral density in the cohort. Finally, it may be argued that the univariate associations of HRV variables were not corrected for multiple testing. We note, however, that the HR for hip fracture risk for SDNN was similar on univariate analysis (0.80) to that when all HRV variables were considered together (0.74).

In conclusion, in this exploratory analysis, increased HRV appears to play an independent protective role for hip fracture risk among older women. Future studies should confirm these findings and examine the mechanisms through which ANS modulation of HRV impacts bone health.