The effect of physical activity on dose-relationship between serum 25-hydroxyvitamin D and cardiovascular health events in older adults

Abstract

Background and aims:

Reverse J- or U-shaped associations between serum 25-hydroxyvitamin D (25[OH]D) concentrations and cardiovascular outcomes have been reported, which need clarifications in older adults. Physical activity, correlating with both serum 25[OH]D concentration and cardiovascular health, may have an effect on the dose-relationships.

Methods:

At baseline, 2790 participants aged 65 years and over, free of vitamin D supplementation use, had assays for serum 25[OH]D concentrations and health related characteristics and measurements, were followed up for cardiovascular events and death by up to 7 and 15 years, respectively. The dose-response associations of serum 25[OH]D concentrations with cardiovascular events and mortality risk were examined using Cox regression models.

Results:

After adjusting for physical activity and other covariates, serum 25[OH]D concentration was non-linearly associated with cardiovascular mortality risk (U-shaped, P = 0.009). According to the Institute of Medicine categories, the HR(95% CI) of cardiovascular mortality risk separately in deficient (< 25nmol/L), inadequate (25 to < 50nmol/L) and potentially harmful (≥ 125nmol/L) level was 1.67 (0.23, 12.01), 1.66 (1.25, 2.20) and 2.21 (0.30, 16.37), respectively. The risk of 25[OH]D inadequacy for cardiovascular mortality was significantly attenuated by increased physical activity, especially leisure activity (P for trend = 0.008 and 0.021, respectively). No significant finding was observed for incident cardiovascular events.

Conclusions:

Both lower and higher serum 25[OH]D concentrations were associated with risk of cardiovascular mortality in Chinese community-dwelling older adults. Physical activity may attenuate the cardiovascular mortality risk of vitamin D inadequacy, but its role in individuals with higher 25[OH]D concentrations remains unclear.

Introduction

Cardiovascular diseases (CVDs) are the leading causes of death in the world [1]. Given the aging of the population, multiple approaches should be explored to achieve a substantial risk reduction in premature death from CVDs [2]. Identifying important modifiable risk factors could largely help formulate public health interventions to reduce cardiovascular events and mortality.

Vitamin D has been suggested to be a benefactor for cardiovascular conditions [3], on the ground of its association with the improvement of endothelial function and glucose homeostasis, attenuation of oxidative stress, inflammatory response and thrombogenesis, and modulation of calcium and lipoproteins metabolisms [4]. Low serum concentrations of 25-hydroxyvitamin D (25[OH]D) have also been consistently associated with risk of cardiovascular disease [5–7] and cardiovascular mortality [7–9] in observational studies. However, recent randomised controlled trials (RCTs) have failed to show the beneficial effect of vitamin D supplementation on the incidence of cardiovascular events [10, 11]. The discrepancy between observational and intervention studies suggest that low 25[OH]D is more likely to be a marker of metabolic imbalance [4]. Meanwhile, growing evidences in cardiac patients and middle aged persons have shown a reverse J- or U-shape association between serum 25[OH]D concentration and incident cardiovascular related health events [8, 12, 13]. And a 3-year RCT indicated that long-term supplementation with moderately high vitamin D doses was associated with a greater need for mechanical circulatory support implants in heart failure patients [14]. These observations may further raise the concern over the limited cardiovascular-beneficial value of large increases in serum vitamin D concentrations via supplementations, especially in the older adults who have a frailer metabolic balance. More studies are required to examine the dose-relationships in old population.

In addition, multiple lifestyle and environmental factors could modify the serum vitamin D status. Besides vitamin D intake by nutrition, physical activity is a major way to achieve higher vitamin D serum levels [15–17]. Physical activity is also an important protective factor against CVD and its caused-mortality [18, 19], potentially through some shared mechanisms with vitamin D including improvements in endothelial function, reducing risk of thrombosis and modulating inflammation process [20]. Serum 25[OH]D and physical activity are each individually associated with cardiovascular outcomes, and a synergistic beneficial effect of them in middle ages on reducing long-term CVDs risk has been suggested recently [16]. However, whether serum 25[OH]D in older adults would associate with the longitudinal risk of cardiovascular events and mortality, independently of or interactively with physical activity, needs further identification.

Thus, this study is aimed to examine a) the dose-response relationships of serum 25[OH]D concentration with risk of incident cardiovascular events and cardiovascular mortality, adjusting for physical activity and other covariates, b) whether serum 25[OH]D concentrations and the levels of physical activity have an interactive effect on the risk of cardiovascular events and cardiovascular mortality, in community-dwelling older adults in Hong Kong.

Material and Methods

Study population

The study population was a subgroup of participants in the Mr. OS and Ms. OS Hong Kong cohort, and the methodology has been described previously [21, 22]. Four thousand community-dwelling Chinese men and women aged 65 years or over were recruited from 2001 to 2003. Those who were unable to walk independently, had bilateral hip replacement, or were not competent to give informed consent were excluded. They were recruited using a stratified sampling method so that approximately 33% would be in each of these age groups: 65–69, 70–74, and 75 or over. A random sub-group (selected by generating random numbers) consisting of 2815 (70%) people (1408 men and 1407 women) had baseline serum 25[OH]D and high-sensitivity C-reactive protein (hs-CRP) measured. Twenty-five participants were excluded then for taking vitamin D supplement at baseline, and 2790 people were included in the present study (see flow chart in Supplemental Figure 1). Written informed consent was obtained from each patient included in the study. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki, and the study has been approved by the Institution’s ethics committee on research on humans.

Assessment of demographic and health related characteristics

The baseline assessment included an interview using a standardized and structured questionnaire. Data on demographics, medical history, medication and supplement use, and lifestyle (smoking and alcohol consumption) were collected. Physical activity level was assessed using the Physical Activity Scale of the Elderly (PASE) [23]. It is a 12-item scale that measures the average number of hours per day spent on leisure, household, and occupational physical activities over the previous 7-day period. Activity weight for each item was determined by the amount of energy expenditure, while score for each item was calculated by multiplying the activity weight by its daily frequency. A summary score of all the items was calculated to reflect the daily physical activity level, and three sub-scores for leisure exercise, household work, and occupation were calculated respectively. Given the advanced age of the study population, sub-score for occupation was very low and did not use alone in the present study.

Body weight (kilograms) was measured using the Physician Beam Balance Scale (Healthometer, IL, USA) with wearing an examination gown. Body height (centimeters) was measured with a Holtain Harpenden stadiometer (Holtain Ltd., Crosswell, UK). Body mass index (BMI) (kg/m²) was calculated. Blood pressure was measured after 5-minute rest in the sitting position using a standard mercury sphygmomanometer (WA Baum Co. Inc., Copiague, NY, USA) by the trained staff. The first and fifth Korotkoff phases were recorded as systolic blood pressure (SBP) and diastolic blood pressure (DBP). The average of two readings was taken. Hypertension was defined as SBP ≥ 140 mmHg or DBP ≥ 90 mmHg, or a self-reported physician diagnosis of hypertension. Blood pressure at the ankle and the blood pressure in the upper arm (brachium) were measured, and Ankle-brachial index (ABI) was calculated [24].

Assessment of bone mineral density (BMD)

Total hip, femoral neck, and lumbar spine areal BMD were measured with DXA using a Hologic QDR 4,500 W device (Waltham, MA, USA). The osteoporosis category was defined as either the T-score of femoral neck BMD, total hip BMD or spine BMD was −2.5 or lower. T-score for femoral neck BMD or total hip BMD was calculated based on the NHANES Ⅲ reference database for corresponding measurements in women aged 20 – 29 years [25]. T-score for lumbar spine BMD was calculated using databases of the manufacturers.

Serum 25[OH]D, hs-CRP, and creatinine assay

Archived baseline serum stored at −80°C for assays. Total 25[OH]D and creatinine were measured by using the liquid chromatography-mass spectrometry (LC-MS) method [26, 27]. Hs-CRP was measured using a commercially available enzyme-linked immunosorbent assay (Vitros Fusion 5,1, Vitros Chemistry Products, USA) and was performed by PathLab Co. Ltd. Coefficient of variation for the analytes was 4 – 7% for 25[OH]D, 3 – 7% for creatinine, and 2 – 5% for hs-CRP. Among the 2790 participants qualified for serum 25[OH]D and hs-CRP assays, 2100 (75%) had an assay for serum creatinine. The estimated glomerular filtration rate (eGFR) was calculated using the Modification of Diet in Renal Disease study equation based on standardized serum creatinine, gender and age [28]. Due to the missing data (24%) for eGFR, eGFR status (defined as normal: ≥ 60; abnormal: < 60; and missing) was used for further adjustments in the multivariate analyses.

Ascertainment of incident cardiovascular events and cardiovascular mortality

Incident cardiovascular events and cardiovascular mortality during the follow-up were the primary endpoints in the present study. The Clinical Management System (CMS) of the Hong Kong Hospital Authority (HA) was used to retrieve retrospectively the hospital discharge diagnosis of cardiovascular events (including ischemic and hemorrhagic stroke, angina, acute myocardial infarction, heart failure, acute coronary syndrome, and transient ischemic attack) from baseline to September 2008. Data on all-cause mortality, from baseline to March 2017, was obtained from Death Registry of the Department of Health of the Hong Kong SAR Government where all deaths in Hong Kong are registered. The total cardiovascular disease mortality was defined according to the International Classification of Diseases, 10^th revision (ICD-10) codes (I00-I99).

Statistical analysis

Baseline characteristics for those had or did not have incident cardiovascular event and cardiovascular mortality were shown as mean (standard deviation, SD) for normally distributed variables, median (interquartile range) for skewed distributed variables and frequency (percent) for categorical variables. The differences across groups were examined by using Student’s t test for normally distributed variables or log-transformed values of variables with skewed distribution, and Chi-Square test for categorical variables. According to the Institute of Medicine (IOM) [29], subjects were grouped by four categories of baseline serum 25[OH]D concentration: < 25nmol/L (deficient), 25to < 50nmol/L (inadequate), 50 to < 125nmol/L (adequate), and ≥125nmol/L (potentially harmful). Baseline demographics, and health related factors and events were shown as mean (SD) for normally distributed variables, median (interquartile range) for skewed distributed variables and frequency (percent) for categorical variables, separately by the four categories. For each factor, P values for difference across categories were calculated using Analysis of Variance (ANOVA) for continuous variables (log-transformed values of variables with skewed distribution) and using Chi-Square test for categorical variables, respectively. Given the small number of subjects (n=3) in the group of 25[OH]D ≥ 125nmol/L, P values for difference across groups without whose 25[OH]D ≥ 125nmol/L were provided additionally.

To test the dose-response association between serum 25[OH]D and cardiovascular outcomes, serum 25[OH]D concentration was entered in the Cox regression model as a restricted cubic spline [30] with 3 knots at the 25nmol/L, 50nmol/L, and 125nmol/L, with a median of 60nmol/L as reference. Cox regression model was also used to separately estimate the hazards ratio (HRs) of IOM serum 25[OH]D categories (adequacy: 50 to < 125nmol/L as reference) for cardiovascular events and cardiovascular mortality risk. Interaction analysis between serum 25[OH]D categories and physical activity levels (the tertiles of PASE score: 1^st tertile (inactive), 2^nd tertile (moderate active), and 3^rd tertile (most active)) was conducted by introducing an interaction term of them in the Cox model. The HRs of serum 25[OH]D categories (the category of serum 25[OH]D adequacy as reference) for cardiovascular events and cardiovascular mortality risk were further examined by the stratification of three physical activity levels. To elucidate the role of physical activity types, the leisure exercise and household work activity were separately considered in the interaction analyses and stratification analyses. The basic models were adjusted for baseline age, gender, and season of blood sampling. And they were additionally adjusted for baseline BMI, physical activity, and bio-medical conditions (cardiovascular diseases, hypertension, diabetes, osteoporosis, ABI, serum level of hs-CRP, and eGFR status) in the full models. Considering the characteristics of the present study (OS cohort), treatment for osteoporosis at baseline and during follow-up, vitamin D treatment during the follow-up, presence of fractures at baseline and occurrence of fractures during the follow-up were additionally adjusted for exploration. 95% confidence interval (CI) was reported for each HR. Participants who reported a history of CVDs were excluded in the sensitivity analyses.

All statistical tests were two-tailed with P < 0.05 being considered as significant. Statistical analyses were performed using SAS 9.4 (SAS Institute, Inc., Cary, NC, USA).

Results

A total of 2790 subjects were included in the present study, with an average age of 72.6 (5.3) years and average serum 25[OH]D concentration of 60.4 (15.3) nmol/L. Three hundred and forty-two (12.3%) subjects had at least one incident cardiovascular event during a median follow-up of 5.7 years (range 0.03 to 7.1 years). Two hundred and thirty-one (11.2%) died from cardiovascular disease during a median follow-up of 13.8 years (range 0.21 to 15.6 years). The demographic characteristics and risk factors compared between those had or did not have incident cardiovascular events and cardiovascular mortality were separately presented in Table 1. Subjects who had an incident cardiovascular event or died from cardiovascular disease were older, and were more likely to be male and less active, have lower ABI, higher serum hs-CRP and lower eGFR, and have cardiovascular disease, diabetes and hypertension at baseline.

Table 1.

The baseline demographic characteristics and risk factors for incident cardiovascular events and mortality risk in old adults.

Variables	Cardiovascular events b		P value c	Cardiovascular disease mortality b		P value c
	No (n=2448)	Yes (n=342)		Survival (n=1839)	Died (n=231)
Age (y)	72.3 (5.1)	74.6 (5.8)	<0.001	71.3 (4.5)	75.6 (5.7)	<0.001
BMI (kg/m2)	23.7 (3.3)	24 (3.3)	0.081	23.9 (3.1)	24 (3.4)	0.584
PASE score	90.9 (42)	86.1 (42)	0.046	94 (42.9)	80.7 (36.7)	<0.001
Serum 25[OH]D (nmol/L)	60.3 (15)	61.4 (17)	0.250	60.6 (14.8)	60.6 (17.4)	0.942
Hs-CRP (mg/L)	1.6 (2.6)	2.1 (2.8)	<0.001	1.6 (2.4)	2.2 (3.1)	<0.001
eGFRa	87.4 (20.8)	82.9 (20.3)	0.003	88.8 (20.4)	84 (20.6)	0.009
Ankle-brachial index	1.06 (0.12)	1.05 (0.16)	0.139	1.07 (0.11)	1.02 (0.17)	<0.001
Gender (female)	1259 (51.4)	134 (39.2)	<0.001	1032 (56.1)	100 (43.3)	<0.001
Less than secondary education (Yes)	1774 (72.5)	249 (72.8)	0.895	1310 (71.2)	164 (71.0)	0.940
Current smoker (Yes)	165 (6.7)	25 (7.3)	0.695	97 (5.3)	9 (3.9)	0.370
Daily alcohol drinking (Yes)	10 (0.4)	0 (0)	0.236	4 (0.2)	1 (0.4)	0.530
Osteoporosis (Yes)	1060 (43.3)	133 (38.9)	0.122	822 (44.7)	88 (38.1)	0.057
Diabetes (Yes)	320 (13.1)	70 (20.5)	<0.001	211 (11.5)	53 (22.9)	<0.001
Hypertension (Yes)	1750 (71.5)	277 (81.0)	<0.001	1323 (71.9)	184 (79.7)	<0.001
CVD (Yes)	433 (17.7)	115 (33.6)	<0.001	323 (17.6)	70 (30.3)	<0.001

BMI: body mass index; 25[OH]D: 25-hydroxyvitamin D; Hs-CRP: high-sensitivity C-reactive protein; eGFR: estimated glomerular filtration rate; PASE:

Physical Activity Scale of the Elderly; CVD: cardiovascular diseases.

^athere were 2100 and 1657 participants had eGFR data for cardiovascular event and cardiovascular mortality analysis, respectively.

^bdescription of mean (SD) for normally distributed variables, median (interquartile range) for skewed distributed variables, and n (%) for categorical variables.

^cusing Student’s t test for normally distributed variables or log-transformed value of variables with skewed distribution, and Chi-Square test for categorical variables.

As shown in Table 2, 16 (0.6%) subjects had deficient 25[OH]D (<25nmol/L), 743 (26.6%) had an inadequate level of 25[OH]D (25 to < 50nmol/L), 2028 (72.7%) had an adequate level of 25[OH]D (50 to < 125nmol/L) and 3 (0.1%) subjects had higher 25[OH]D greater than 125nmol/L. Subjects with serum 25[OH]D deficient and inadequate were more likely to be female and less active, and they tended to have osteoporosis, lower ABI and higher eGFR (2100 participants had eGFR data for analysis) at baseline. Compared with those had inadequate and higher 25[OH]D, subjects had an adequate level of 25[OH]D at baseline were less likely to have cardiovascular mortality (n=153, 10.1%) during follow-up (P = 0.046).

Table 2.

Baseline characteristics and cardiovascular related events by groups of serum 25[OH]D concentrations.

		Serum 25[OH]D				P Value c/d
	All	< 25 nmol/L	25 to < 50 nmol/L	50 to < 125 nmol/L	≥ 125 nmol/L
	N=2790 (100)	N=16 (0.6)	N=743 (26.6)	N=2028 (72.7)	N=3 (0.1)
Age (y)	72.6 (5.3)	71.7 (4.1)	72.8 (5.6)	72.5 (5.1)	74 (2.6)	0.336/0.206
BMI (kg/m2)	23.7 (3.3)	23.9 (4)	23.9 (3.6)	23.7 (3.2)	23.6 (3.1)	0.492/0.301
PASE score	90.3 (42.0)	82.4 (44.7)	83.7 (39.3)	92.7 (42.5)	167.5 (87.6)	<0.001/<0.001
Serum 25[OH]D (nmolL)	60.4 (15.3)	21.8 (4.1)	43.3 (5.9)	66.9 (11.6)	153.1 (11.3)	<0.001/<0.001
hs-CRP (mg/L)	1.7 (2.7)	2.1 (4.3)	1.6 (2.6)	1.7 (2.8)	2.5 (2.2)	0.847/0.698
eGFRa	87 (20.8)	91.6 (24.3)	91.5 (21.2)	85.3 (20.3)	-	−/<0.001
Ankle-brachial index	1.06 (0.13)	1.01(0.1)	1.05 (0.13)	1.07 (0.12)	1.09 (0.02)	0.029/0.012
Blood sampling in autumn/winter	1532 (54.9)	10 (62.5)	423 (56.9)	1099 (54.2)	0 (0)	0.129/0.365
Gender (female)	1393 (49.9)	12 (75.0)	463 (62.3)	918 (45.3)	0 (0)	<0.001/<0.001
Less than secondary education (Yes)	2023 (72.5)	10 (62.5)	522 (70.3)	1489 (73.4)	2 (66.7)	0.309/0.170
Current smoker (Yes)	190 (6.8)	1 (6.3)	48 (6.5)	141 (7.0)	0 (0)	0.933/0.898
Daily alcohol drinking (Yes)	10 (0.4)	0 (0)	1 (0.1)	9 (0.4)	0 (0)	0.676/0.469
Osteoporosis (Yes)	1193 (42.8)	11 (68.8)	380 (51.1)	801 (39.5)	1 (33.3)	<0.001/<0.001
Diabetes (Yes)	390 (14.0)	4 (25.0)	106 (14.3)	279 (13.8)	1 (33.3)	0.443/0.417
Hypertension (Yes)	2027 (72.7)	12 (75.0)	538 (72.4)	1476 (72.8)	1 (33.3)	0.490/0.960
CVD (Yes)	548 (19.6)	2 (12.5)	145 (19.5)	400 (19.7)	1 (33.3)	0.828/0.766
Incident events
Incident CVD event (Yes)	342 (12.3)	2 (12.5)	93 (12.5)	246 (12.1)	1 (33.3)	0.725/0.962
CVD death (Yes)	231 (11.2)	1 (10.0)	76 (14.1)	153 (10.1)	1 (33.3)	0.046/0.039

BMI: body mass index; 25[OH]D: 25-hydroxyvitamin D; Hs-CRP: high-sensitivity C-reactive protein; eGFR: estimated glomerular filtration rate; PASE:

Physical Activity Scale of the Elderly; CVD: cardiovascular disease.

^aThere were 2100 participants had eGFR data for analysis.

^bdescription of mean (SD) for normally distributed variables, median (interquartile range) for skewed distributed variables, and n (%) for categorical variables.

^cP value for difference across vitamin D groups for each measure was calculated using Analysis of Variance (ANOVA) for normally distributed variables or log-transformed value of variables with skewed distribution, and using Chi-Square test for categorical variables.

^dP value for difference across groups without whose 25[OH]D ≥ 125 nmol/L.

In the full model, the dose-response analyses of restricted cubic spline Cox regression showed a marginally significant non-linear association of serum 25[OH]D concentration with incident cardiovascular events and a significant non-linear association with cardiovascular disease mortality risk, respectively (P = 0.082 and 0.009, respectively; Figure 1 and Table 3). In the Cox model, compared with subjects who had an adequate level of serum 25[OH]D, the HRs (95%CI) of those having deficient, inadequate and higher 25[OH]D for incident cardiovascular events risk were 1.45 (0.36, 5.85), 1.08 (0.85, 1.38) and 2.73 (0.38, 19.51) in the basic models, respectively. The HRs (95%CI) for cardiovascular mortality were 1.76 (0.25, 12.71), 1.61 (1.22, 2.13) and 2.29 (0.32, 16.47), respectively. After full adjustments for baseline age, gender, season of blood sampling, BMI, physical activity, and baseline bio-medical conditions (cardiovascular diseases, hypertension, diabetes, osteoporosis, ABI, serum level of hs-CRP, and eGFR status), or and additional adjustments for treatment for osteoporosis at baseline and during follow-up, vitamin D treatment during the follow-up, presence of fractures at baseline and occurrence of fractures during the follow-up, no essential change for the above estimates had been shown. (Table 3)

Figure 1.

Hazards ratios of serum 25[OH]D concentrations for risk of incident cardiovascular events (A) and cardiovascular mortality (B) by restricted cubic spline Cox regression analysis with adjustments for baseline age, gender, and season of blood sampling, body mass index, physical activity and baseline bio-medical conditions (cardiovascular diseases, hypertension, diabetes, osteoporosis, ankle-brachial index, serum level of hs-CRP and eGFR status). Knots were placed on the 25, 50, and 125 nmol/L of the serum 25[OH]D concentration. The 60 nmol/L (median) was used as the reference for incident cardiovascular events and cardiovascular mortality risk.

Table 3.

The associations of serum 25[OH]D group with the incident cardiovascular events and mortality.

	Cardiovascular events, HR (95%CI)
	25[OH]D < 25 nmol/L	25[OH]D 25 to < 50 nmol/L	25[OH]D 50 to < 125 nmol/L	25[OH]D ≥ 125 nmol/L
Models	N=2/16	N=93/743	N=246/2028	N=1/3
Basic model a	1.45 (0.36, 5.85)	1.08 (0.85, 1.38)	Reference	2.73 (0.38, 19.51)
Full model b	1.61 (0.40, 6.53)	1.08 (0.85, 1.39)	Reference	1.91 (0.25, 14.37)
Additionally adjusted model c	1.61 (0.40, 6.51)	1.07 (0.84, 1.37)	Reference	1.79 (0.24, 13.52)
P value for non-linearity	0.022a/0.082b/0.092c
	Cardiovascular mortality, HR (95%CI)
	25[OH]D < 25 nmol/L	25[OH]D 25 to < 50 nmol/L	25[OH]D 50 to < 125 nmol/L	25[OH]D ≥ 125 nmol/L
	N=1/10	N=76/539	N=153/1518	N=1/3
Basic model a	1.76 (0.25, 12.71)	1.61 (1.22, 2.13)	Reference	2.29 (0.32, 16.47)
Full model b	1.67 (0.23, 12.01)	1.66 (1.25, 2.20)	Reference	2.21 (0.30, 16.37)
Additionally adjusted model c	1.67 (0.23, 12.05)	1.66 (1.25, 2.21)	Reference	2.62 (0.34, 19.84)
P value for non-linearity	0.004a/0.009b/0.007c

^aadjusting for baseline age, gender, and season of blood sampling;

^badjusting for baseline age, gender, season of blood sampling, body mass index, physical activity, and baseline bio-medical conditions (cardiovascular diseases, hypertension, diabetes, osteoporosis, ankle-brachial index, serum level of hs-CRP and eGFR status);

^cadjusting for baseline age, gender, season of blood sampling, body mass index, physical activity, and baseline bio-medical conditions (cardiovascular diseases, hypertension, diabetes, osteoporosis, ankle-brachial index, serum level of hs-CRP and eGFR status), treatment for osteoporosis at baseline and during follow-up, vitamin D supplements during follow-up, presence of bone fractures at baseline and occurrence of fractures during follow-up.

A significant interaction between categories of serum 25[OH]D and physical activity levels for cardiovascular mortality risk was shown (P for interaction = 0.017, 0.031 and 0.025 in the basic, full and additionally adjusted model, respectively). Compared with those had adequate serum 25[OH]D, the adjusted HRs (95%CI) of having inadequate serum 25[OH]D for cardiovascular mortality risk were 2.29 (1.53, 3.43), 1.19 (0.72, 1.99) and 0.90 (0.46, 1.78) respectively in those being inactive, moderate active and most active (P for trend = 0.008 in the basic model). Compared with those had adequate serum 25[OH]D, the adjusted HRs (95%CI) of having deficient 25[OH]D for cardiovascular mortality risk was 6.57 (0.89, 48.79) in those being inactive, and of having higher 25[OH]D for cardiovascular mortality risk was 6.09 (0.82, 45.13) in those being most active. Test for trend failed to conduct in these two groups due to limited numbers of subjects. In addition, a significant interaction between categories of serum 25[OH]D and leisure-exercise physical activity levels (P for interaction = 0.004, Table 5), but not household work activity levels (P for interaction = 0.158), for cardiovascular mortality risk was shown. The estimates above were not essentially changed in the full and additionally adjusted models. No significant interactive effect of serum 25[OH]D and physical activity levels was shown for the risk of cardiovascular events.

Table 5.

The associations of serum 25[OH]D group with the incident cardiovascular mortality by level of leisure physical activity.

Leisure physical activity levels	Cardiovascular mortality, HR (95%CI)
	25[OH]D	25[OH]D	25[OH]D	25[OH]D
	< 25 nmol/L	25 to < 50 nmol/L	50 to < 125 nmol/L	≥ 125 nmol/L
Inactive	N=1/4	N=39/211	N=42/442	N=0/1
Basic model a	6.57 (0.89, 48.79)	2.78 (1.78, 4.35)	Reference	-
Full model b	5.92 (0.78, 44.73)	3.14 (1.98, 4.98)	Reference	-
Additionally adjusted model c	5.84 (0.77, 44.49)	3.17 (2.00, 5.02)	Reference	-
Moderate active	N=0/5	N=23/186	N=54/513	N=0/0
Basic model a	-	1.22 (0.75, 2.00)	Reference	-
Full model b	-	1.38 (0.83, 2.31)	Reference	-
Additionally adjusted model c	-	1.39 (0.82, 2.35)	Reference	-
Most active	N=0/1	N=14/142	N=57/563	N=1/2
Basic model a	-	0.91 (0.49, 1.69)	Reference	6.09 (0.82, 45.13)
Full model b	-	0.98 (0.54, 1.81)	Reference	3.68 (0.45, 30.09)
Additionally adjusted model c	-	0.98 (0.53, 1.83)	Reference	3.56 (0.43, 29.58)
P for trend	-	0.021a/0.028b/0.033c	0.276a/0.292b/0.199c	-
P for interaction		0.004a/0.004b/0.004c

^aadjusting for baseline age, gender, and season of blood sampling;

^badjusting for baseline age, gender, season of blood sampling, body mass index and baseline bio-medical conditions (cardiovascular diseases, hypertension, diabetes, osteoporosis, ankle-brachial index, serum level of hs-CRP and eGFR status);

^cadjusting for baseline age, gender, season of blood sampling, body mass index, physical activity, and baseline bio-medical conditions (cardiovascular diseases, hypertension, diabetes, osteoporosis, ankle-brachial index, serum level of hs-CRP and eGFR status), treatment for osteoporosis at baseline and during follow-up, vitamin D supplements during follow-up, presence of bone fractures at baseline and occurrence of fractures during follow-up.

Excluding participants who had reported CVDs at baseline attenuated some but did not essentially change the results in the sensitivity analyses.

Discussion

The present study observed a significant non-linear association between serum 25[OH]D concentration and risk of cardiovascular mortality in Chinese community-dwelling older adults, with a median follow-up of 13.8 years. According to the IOM categories, a potential increasing risk was observed at both a lower and higher level than adequate 25[OH]D (50 to < 125nmol/L). Further adjusting for physical activity levels and other covariates did not significantly change the associations. The association between serum 25[OH]D inadequacy and higher cardiovascular mortality risk was significantly different across the levels of physical activity. In older adults who had higher physical activity levels (especially higher leisure exercise levels), those who had inadequate 25[OH]D were less likely to be associated with cardiovascular mortality risk. Numbers of subjects in the groups of 25[OH]D deficiency and greater than 125nmol/L were limited to observed their varied associations with cardiovascular mortality across physical activity levels. No significant finding was observed for the risk of cardiovascular events.

Consistent with two previous studies conducted in middle aged persons [8, 13], both lower and higher serum 25[OH]D were potentially associated with a higher risk of cardiovascular outcomes in older adults. Especially, we observed the increased risk in relation to serum 25[OH]D inadequacy was stronger for cardiovascular mortality in a long-term view than for cardiovascular events in a shorter term. And if follow-up time for cardiovascular mortality was truncated to 7 years, no significant association was observed (data not shown). This is partly in line with the findings from the RCTs that vitamin D supplementation did not reduce the cardiovascular related health events in a follow-up time less than 10 years [11, 31]. Whereas, due to the small numbers of two 25[OH]D groups with deficient (<25nmol/L) and higher (≥125nmol/L) values in this relative healthy old population may limited the statistical power to show the significance, only a moderately but non-significantly increased risk in relation to the categories of deficient and higher serum 25[OH]D for cardiovascular events and cardiovascular mortality was observed. Nevertheless, results from the dos-response analyses using restricted cubic spline functions (based on assumed stronger theoretical advantages than using categorization) [30] demonstrated the significant non-linear association between serum 25[OH]D and cardiovascular mortality risk.

Physical activity, as aforementioned, contributes largely to increases in serum vitamin D concentrations [15, 16]. Although it is an important protector for reducing cardiovascular related outcomes [18, 32], exercise-associated acute cardiovascular events and sudden cardiac death also non-negligibly take place in susceptible persons [33, 34]. These make physical activity potentially play a part in the dose-relationship between serum 25[OH]D concentration and cardiovascular health. However, further adjusting for physical activity levels did not alter the general association between elevated serum 25[OH]D concentration and cardiovascular events or mortality risk in the present study. The association between serum 25[OH]D inadequacy and cardiovascular mortality risk was indeed attenuated with the increasing levels of physical activity (especially the leisure ones). But larger sample sizes are needed to confirm the role of physical activity for the categories of 25[OH]D less than 25 nmol/L and greater than 125 nmol/L. Higher risk for cardiovascular mortality in older adults had inadequate serum 25[OH]D may be attenuated through either increasing serum 25[OH]D concentrations or engaging more activities only, or increasing them both. These indicates that serum vitamin D and physical activity appear to additionally protect against cardiovascular mortality, aside from their shared mechanisms beneficial for cardiovascular conditions [4, 35]. In addition, previous studies suggested that increasing physical activity was not an equally effective way for improving 25[OH]D deficiency among races [16, 36], especially greater skin pigmentation tends to have less vitamin D production [37]. Thus, our findings greatly support the efforts to increase physical activity in older adults, not restricted to raising serum vitamin D, to reduce the cardiovascular mortality risk of vitamin D inadequacy.

Besides, vascular calcification and bone mineralization have been suggested to play a part in the association of higher vitamin D with the risk of cardiovascular event [38, 39]. However, further adjusting for serum calcium, phosphate and parathyroid hormone (PTH) concentrations [13], or for the presence of osteoporosis and related treatment, the presence and occurrence of fractures, and use of vitamin D supplementations during follow-up as in the present study did not alter the association between serum 25[OH]D and cardiovascular related health events. Further, a recent trial in swine found that high calcium diets (up to 2,000 milligrams a day) had no detectable effect on coronary artery calcium deposition [40], and vitamin D intakes reaching at an extremely high serum 25[OH]D concentration (such as > 375nmol/L) can be detrimental [41]. Renal function may also play a role, as kidneys are involved in the metabolism of vitamin D [38, 42]. Although a close association of kidney disease with hypovitaminosis D has been suggested [43, 44], in people without chronic kidney disease, the high concentration of inactive forms of vitamin D was found to be associated with the reduction of estimated renal function [45]. Our subjects with higher serum 25[OH]D did indeed have lower eGFR, but the associations between serum 25[OH]D and risk of cardiovascular events or mortality were not significantly changed by adjusting for eGFR or varied across eGFR status. Further studies are required to elucidate the risk of largely elevated serum 25[OH]D, such as greater than 125 nmol/L.

The strengths of the study include the relative large sample of well-characterized community-dwelling older participants with up to 7 years and 15 years of follow-up for cardiovascular events and death, respectively. Cardiovascular events and cardiovascular death were prospectively and systematically identified from well-maintained government institutions. The limitations included that there was only a single measurement of serum 25[OH]D and that serum calcium, phosphate, and PTH were not available. In addition, a relative shorter term follow-up for cardiovascular events (due to limited accessibility to HA data), and small sample size in the stratified analysis limited the power to find the consistent associations. As in all observational studies, no causality between serum vitamin D concentration and cardiovascular events and mortality risk could be implied.

In conclusion, serum 25[OH]D status was non-linearly associated with the risk of cardiovascular mortality in the Chinese community-dwelling older adults. These associations were independently of physical activity levels and other co-variates. Higher risk for cardiovascular mortality in older adults had inadequate serum 25[OH]D may be attenuated through either increasing serum 25[OH]D concentrations or engaging more activities (especially the leisure exercise). More studies are required to clarify the underlying mechanisms for the increased risk of cardiovascular related health events in relation to higher elevated serum 25[OH]D and its interrelationship with physical activity.

Table 4.

The associations of serum 25[OH]D group with the incident cardiovascular mortality by level of physical activity.

Physical activity levels	Cardiovascular mortality, HR (95%CI)
	25[OH]D	25[OH]D	25[OH]D	25[OH]D
	< 25 nmol/L	25 to < 50 nmol/L	50 to < 125 nmol/L	≥ 125 nmol/L
Inactive	N=1/4	N=43/194	N=53/459	N=0/1
Basic model a	4.38 (0.59, 32.43)	2.29 (1.53, 3.43)	Reference	-
Full model b	5.57 (0.74, 41.86)	2.34 (1.55, 3.53)	Reference	-
Additionally adjusted model c	5.40 (0.72, 40.68)	2.36 (1.56, 3.57)	Reference	-
Moderate active	N=0/3	N=21/189	N=54/489	N=0/0
Basic model a	-	1.19 (0.72, 1.99)	Reference	-
Full model b	-	1.31 (0.78, 2.22)	Reference	-
Additionally adjusted model c	-	1.26 (0.74, 2.13)	Reference	-
Most active	N=0/3	N=12/156	N=46/570	N=1/2
Basic model a	-	0.90 (0.46, 1.78)	Reference	6.60 (0.89, 49.14)
Full model b	-	0.98 (0.49, 1.94)	Reference	4.03 (0.49, 33.35)
Additionally adjusted model c	-	0.96 (0.48, 1.92)	Reference	4.30 (0.51, 36.47)
P for trend	-	0.008a/0.020b/0.018c	0.468a/0.279b/0.421c	-
P for interaction		0.017a/0.031b/0.025c

^aadjusting for baseline age, gender, and season of blood sampling;

^badjusting for baseline age, gender, season of blood sampling, body mass index and baseline bio-medical conditions (cardiovascular diseases, hypertension, diabetes, osteoporosis, ankle-brachial index, serum level of hs-CRP and eGFR status);

^cadjusting for baseline age, gender, season of blood sampling, body mass index, physical activity, and baseline bio-medical conditions (cardiovascular diseases, hypertension, diabetes, osteoporosis, ankle-brachial index, serum level of hs-CRP and eGFR status), treatment for osteoporosis at baseline and during follow-up, vitamin D supplements during follow-up, presence of bone fractures at baseline and occurrence of fractures during follow-up.

Highlights.

• Serum 25[OH]D was non-linearly associated with cardiovascular mortality in older adults.

• This non-linear association was independent of physical activity levels.

• Risk of inadequate 25[OH]D for cardiovascular mortality varied by activity levels.