Risk of Heart Failure Among Postmenopausal Women: A Secondary Analysis of the Randomized Trial of Vitamin D Plus Calcium of the Women’s Health Initiative

Abstract

Background

Vitamin D supplementation may be an inexpensive intervention to reduce heart failure (HF) incidence. However, there are insufficient data to support this hypothesis. This study evaluates whether vitamin D plus calcium (CaD) supplementation is associated with lower rates of HF in post-menopausal women and whether the effects differ between those at high versus low risk for HF.

Methods and Results

Analyses were restricted to 35,983 (of original 36,282) women aged 50 to 79 years old in the Women’s Health Initiative randomized trial of CaD supplementation who were randomized 1:1 in a double-blinded fashion to receive 1,000 mg/day of calcium plus 400 IU/day of vitamin D₃ or placebo. Overall, 744 adjudicated incident HF cases (intervention, 363; control, 381) occurred during a median follow up of 7.1 (interquartile range, 1.6) years. CaD supplementation, compared to placebo, was not associated with reduced HF risk in the overall population, hazard ratio (HR), 0.95; P=0.46. However, CaD supplementation had differential effects (P-interaction=0.005) in subgroups stratified by baseline risk status of HF defined by the presence (high-risk=17,449) or absence (low-risk=18,534) of preexisting HF precursors including coronary heart diseases, diabetes, or hypertension: 37% (HR, 0.63 [95% CI, 0.46 to 0.87]) lower risk of HF in the low-risk versus HR, 1.06; P=0.51, in the high-risk subgroups.

Conclusions

CaD supplementation did not significantly reduce HF incidence in the overall cohort, however, it was beneficial among postmenopausal women without major HF precursors while of little value in high-risk subgroups. Additional studies are warranted to confirm these findings and investigate the underlying mechanism.

Heart failure (HF) is a major public health problem with an enormous burden of morbidity, disability, and associated health care costs¹. HF incidence rates are likely to remain high due to a rapidly aging population and a high prevalence of major precursors of HF, including hypertension, diabetes, and coronary heart disease (CHD). It is therefore imperative to identify effective interventions for prevention of HF.

The role of vitamin D in HF pathogenesis in humans is not well established, partly owing to the lack of randomized controlled trial data on this relationship. However, it has been proposed that vitamin D can directly affect cardiac functioning through interaction with cardiomyocyte vitamin D receptors (VDR)^{2, 3} or indirectly by modifying the incidence of hypertension, diabetes, CHD, or other major precursors of HF⁴. Evidence for this relationship derives from trials in HF and chronic kidney disease (CKD) patients^5–10. Among CKD patients, it has been suggested that insufficient vitamin D may lead to cardiac malfunction in the presence of excess parathyroid hormone (PTH) levels precipitated by low vitamin D levels^{5, 7–10}. It has also been reported that supplementation with vitamin D has modest effects on ventricular function and N-terminal brain natriuretic peptide (NT-pro BNP) levels among HF patients⁶.

The role of calcium in HF pathogeneses is not clear. There are no existing data on the direct effect of calcium supplementation on HF. However, results from two meta-analyses suggest calcium may be associated with increased risk of some cardiovascular diseases (CVD). Bolland et al¹¹ reported calcium (alone) supplementation to be associated with a significant 27% increased risk of myocardial infarction in a pooled sample of 12,000 trial participants, while Wang et al¹² reported a 14% elevated risk of a composite of CVD associated with calcium (alone) supplementation. However, combined supplementation of calcium and vitamin D had a null effect on CVD incidence¹².

Hsia et al (2007) previously reported that vitamin D plus calcium (CaD) supplementation did not increase the risk of CVD, including HF, in the Women’s Health Initiative (WHI) among CaD trial population¹³. Their analyses involved participants with preexisting HF at baseline; hence, it is not clear what the effect of the intervention was on women free of HF at baseline. The present analyses therefore evaluate the effect of randomized assignment to CaD supplementation in the prevention of HF among post-menopausal women free of HF and participating in the Women’s Health Initiative (WHI) CaD trial. Secondary pre-specified goals for the present analyses were to examine the results in high-risk versus low-risk groups and according to baseline intake of personal vitamin D and calcium supplements. Hypertension, diabetes, and CVD, associated with HF population attributable risks of 59%, 12%, and 26%, respectively, among U.S women aged 40 to 80 years old¹⁴ are also associated with low serum vitamin D levels^{6, 12, 14–17}. Hence vitamin D supplementation may have HF benefits in populations with these preexisting chronic conditions. Pre-specified sensitivity analysis included per-protocol analysis and estimation of CaD effects through inverse probability of censored weights (IPCW) methods¹⁸.

Methods

Design Overview: The Calcium and Vitamin D (CaD) Trial

This is a secondary analysis of HF outcomes from a previously completed RCT testing the effects of CaD on fractures and cancer outcomes as primary endpoints, however, CVD including HF were also ascertained. The CaD trial design and methodology has been published^{19, 20}. This study was approved by an institutional review committee and all subjects gave informed consent.

Randomization and Intervention

The 36,282 postmenopausal women, aged 50 – 79 years old, who met the inclusion criteria and consented to participate in the CaD trial were randomized in a 1:1 ratio double-blinded fashion to receive either the intervention or placebo. The intervention group received a total dosage of 1,000 mg elemental calcium and 400 IU vitamin D₃ per day^{19, 20}. This dose, according to the IOM, was sufficient to meet the recommended dietary allowance (RDA) of both vitamin D and calcium in postmenopausal women. Participants randomized to the placebo group were provided ‘inactive’ tablets that looked exactly like the ‘active’ tablets and instructed to take these tablets in the same format as those in the intervention group to preserve the blinding of the randomization to both study staff and participants¹⁹.

Participants who reported use of personal vitamin D (48.4%) and/or calcium (54.9%) supplements were allowed to continue taking them even after enrolling in the CaD trial. The personal vitamin D supplement consumption was later raised to an upper limit of 1,000 IU/day after the IOM published new recommendations on vitamin D intake in 1999²¹. Details of participant selection for the CaD trial and the present analyses are reported in Figure 1.

Figure 1.

Participant recruitment and selection for the CaD trial and analyses.

Outcome: Ascertainment of Heart Failure Cases

Trained staff abstracted medical records annually for self-reported HF hospitalization and any two-day hospitalization that reported a CVD outcome. The abstracted HF cases were then classified by an adjudication committee of physicians at local clinical sites; only a subset of HF cases were adjudicated by a central adjudication physician committee since HF was a secondary endpoint for the CAD trial. The local and central adjudication methods had an excellent 79% agreement rate (k)²². The local physician committee used the following method: Hospitalized HF (HF requiring and/or occurring during hospitalization) required physician diagnosis of new-onset or worsened congestive HF on the reported hospital admission and 1 or more of the following 4 criteria: 1) HF diagnosed by physician and receiving medical treatment for HF; 2) above plus documentation in the current medical record of a history of an imaging procedure showing impaired left ventricular (LV) systolic or diastolic function; 3) pulmonary edema/congestion on chest X-ray on the current admission; 4) dilated ventricle(s) or “poor” LV or right ventricular (RV) function by echocardiogram, multi gated acquisition (MUGA), radionuclide ventriculogram (RVG), or other contrast ventriculography, or evidence of LV diastolic dysfunction.

Covariates and Effect Modifiers

We stratified the study sample into two subgroups by baseline status of medical histories of major cardiovascular precursors of HF to estimate the effect of CaD intervention in these subgroups. For the pre-specified subgroup analyses, we defined “high-risk” using American College of Cardiology criteria, based on the presence of hypertension, diabetes, CHD or CVD²³.Women who reported any histories of medically diagnosed hypertension, diabetes, CHD, or CVD before randomization were grouped as high-risk of HF while those without any of these conditions were considered to be at low-risk of HF.

Self-reported histories of physician-diagnosed cardiovascular conditions were obtained from all participants during trial enrollment. Details of the description of these baseline factors and other baseline potential confounders listed in Table 1 have been previously reported¹⁹. Hypertension was defined as a self-reported physician diagnosis of hypertension with or without current use of antihypertensive medications or high blood pressure (BP) (measured systolic BP ≥ 140 mmHg and/or diastolic BP ≥ 90 mmHg) during enrollment into the CaD trial. CVD was defined as self-reported physician diagnosis of problems with the heart, blood circulation or clots. CHD was defined as a composite of self-reported cardiac arrest, angina, coronary artery bypass graft surgery (CABG), and/or percutaneous transluminal coronary angioplasty (PTCA). Diabetes was defined based on self-reported physician diagnosis of diabetes and current prescription of diabetic medications.

Table 1.

Baseline Characteristics of Participants Stratified by Baseline Risk Status of Heart Failure and by Treatment Group - the Vitamin D and Calcium (CaD) Trial of the Women’s Health Initiative (WHI) Study, 1995 – 2005

Baseline characteristics	Baseline Risk Status
	Low-risk group (N = 18,534)		High-risk group (N = 17,449)
	CaD (n = 9,307)	Placebo (n = 9,227)	CaD (n = 8,716)	Placebo (n = 8,733)
No. (%) with data
Age (years)
49 – 59	44	44	30	30
60 – 69	43	43	48	48
70 – 81	13	13	22	22
Race/Ethnicity
White	86	87	80	80
Black	6	6	12	12
Hispanic	5	4	4	4
Income ($)
<25,000	18	18	24	25
25, 000 – 50,000	43	44	47	46
50,000 – 70,000	21	20	17	18
75,000 – 100,000	10	10	6	7
>100,000	9	9	5	6
Physical activity level (total METs/week)
Low	37	37	40	40
Moderate	12	13	14	13
High	51	50	47	47
Smoking status
None	52	53	53	54
Past	40	39	40	40
Current	8	8	7	7
Family history of CVD	62	62	70	70
Personal calcium (alone) supplements	50	50	50	50
Personal vitamin D (alone) supplements	50	50	50	50

Abbreviations: SD, standard deviation; BMI, body mass index; MET, metabolic equivalent of task; CVD, cardiovascular disease.

Sample Size and Power Calculations

Women (153 intervention; 146 placebo) with prior medical diagnosis of HF at trial enrollment were excluded from this analysis to create a primary prevention cohort of 35,983 (18,023 intervention; 17,960 control) post-menopausal women free of HF. When stratified by baseline risk status of HF, 17,449 (8,716 intervention; 8,733 control) were considered to be at high-risk while 18,534 (9,307 intervention; 9,227 control) were considered to be at low-risk. A total of 744 HF cases (high-risk: 587 [302 intervention; 285 control] vs low-risk subgroup: 157 [61 intervention; 96 control]) during a total follow up of about 9.8 years. Based on these data and an alpha level of 0.05, this study had 80% statistical power to detect protective effects (hazard ratios) of CaD supplementation of at least 0.86, 0.64, and 0.78 in the overall CaD cohort, the low-risk and high-risk subgroups respectively²⁴. Given a two-tailed a priori hypothesis testing, the study was powered at 80%, type I error of 0.05, to detect interaction effect sizes of ≥ 1.52 when CaD effect is stronger in the high-risk compared to the low-risk group or ≤ 0.66 when CaD effect is stronger in the low-risk compared to the high-risk group²⁵.

Statistical Analysis

All main analyses were performed based on an intention-to-treat approach. Chi-square and Student’s t-test were performed to assess the balance of potential confounders between trial arms for the entire CaD cohort and by stratified subgroups. Cox proportional hazard (CPH) regression models were used to estimate the effect (hazard ratios, HR) of the intervention on HF. Both graphical, Kaplan-Meier curves (Figure 2) and Schoenfeld residuals plots (Supplemental Figure 1) and time-dependent proportionality tests (study cohort [P for interaction]; overall cohort (0.45), low-risk [0.80], high-risk [0.44]) were used to evaluate whether the proportionality assumption was violated by the intervention variable. Formal test of interaction between the randomization status and a binary indicator of baseline HF risk status was performed by including a product term between the two variables in the CPH models. All HRs were estimated from unadjusted CPH models since all potential covariates evaluated were balanced between the two arms of the study.

Figure 2.

Kaplan-Meier curves comparing the cumulative incidence of HF between the CaD and placebo arms during follow-up period in the overall CaD cohort (A) and stratified baseline subgroups (B).

We also evaluated whether personal consumption of vitamin D or calcium supplements at baseline modified the effect of CaD on HF incidence since participants were allowed to continue consuming their personal calcium and vitamin D supplements after enrolling in the CaD trial. Additionally, we tested whether total (diet plus supplements) vitamin D or calcium intake modified the effect of CaD on HF incidence. We also estimated a potential effect modification by self-reported postmenopausal hormone therapy and randomization to receive intervention in the hormone replacement therapy (HRT) trial.

Sensitivity analyses were performed by restricting the analyses to only participants who achieved 80% adherence rate to study medication (n = 23,601, 65.6%). To estimate CaD effects (HRs) independent of censoring information, the IPCW method was implemented^{18, 26}. The IPCW model was adjusted for some of the factors previously reported as strong predictors of adherence to study medication in the CaD trial – age, education level, use of personal calcium, vitamin D or multivitamin supplements, history of HF risk factors, family history of CVD, and enrollment in other clinical trials²⁷.

Results

Baseline socio-demographic, physical/lifestyle, and clinical factors were proportionally distributed between the two study arms for the entire CaD cohort (Supplemental Table 1) and in the two stratified subgroups of participants with and without major precursors of HF (Supplemental Table 2). Baseline known CVD risk factors were more prevalent in the high-risk group compared to the low-risk group, however, personal calcium and vitamin D supplements consumption was proportional between these subgroups (Table 1). There were 744 HF cases (29.0/10,000 person-years) during a median follow-up years of 7.06 (interquartile range: 1.61); 363 (28.2/10,000 person-years) of these occurred in the intervention arm versus 381 (29.8/10,000 person-years) in the placebo arm. When stratified by baseline risk status, more HF cases occurred in the high-risk subgroup (587 [302 intervention; 285 control]) than in the low-risk subgroup of women (157 [61 intervention; 96 control]), P value of difference <0.001.

Supplementation with CaD was not associated with risk of HF hospitalization in the overall cohort, HR, 0.95 [95% CI, 0.82 to 1.09]; P = 0.46. The effect of CaD, however, was modified (P for interaction = 0.005) by baseline risk status of HF (defined by the presence or absence of CHD, other CVD, hypertension, or diabetes) at baseline. When stratified by baseline risk status, CaD was associated with a statistically significant 37% lower risk of HF, HR, 0.63 [95% CI, 0.46 to 0.87]; P = 0.005 (number needed-to-treat, 255 [95%CI, 174 to 723] in the low-risk subgroup, but not in the high-risk subgroup (HR, 1.06 [95% CI, 0.90 to 1.24]) (Figure 3). Consumption of personal vitamin D (P for interaction = 0.33) or calcium (P for interaction = 0.11) supplements did not modify the effect of CaD on HF incidence. The effects of CaD were also not modified by baseline self-reported total (diet plus supplements) consumption of vitamin D (P for interaction = 0.26) or calcium (P for interaction = 0.81) (Table 2). Neither assignment to receive the intervention (estrogen alone or estrogen plus progestin) nor self-reported use of postmenopausal hormone therapy at baseline modified the effect of CaD on HF incidence (Table 2). Baseline covariates also did not confound these associations in multivariable-adjusted CPH regression models.

Figure 3.

Forest plot of hazard ratios for effect of vitamin D plus calcium supplementation on HF.

Table 2.

Effects of the intervention stratified by baseline supplements (calcium and vitamin D) and postmenopausal hormone therapy use

Baseline supplement intake	Overall		Low-risk		High-risk		P - interaction
	95% CI	P	95% CI	P	95% CI	P
Non-protocol calcium supplements							0.11
No (n = 16,218)	1.07 (0.87 – 1.31)	0.54	0.73 (0.46 – 1.16)	0.18	1.19 (0.95 – 1.50)	0.14
Yes (n = 19,765)	0.85 (0.69 – 1.03)	0.102	0.55 (0.36 – 0.87)	0.01	0.94 (0.75 – 1.18)	0.58
Non-protocol vitamin D supplements							0.33
No (n = 18,558)	1.01 (0.83 – 1.23)	0.91	0.74 (0.47 – 1.14)	0.17	1.12 (0.90 – 1.39)	0.33
Yes (n = 17,425)	0.88 (0.71 – 1.08)	0.22	0.53 (0.33 – 0.85)	0.01	0.99 (0.78 – 1.26)	0.93
Hormone therapy trial assignment							0.83
Not randomized or randomized to receive placebo (n=27,929)	0.94 (0.80 – 1.11)	0.45	0.76 (0.52 – 1.09)	0.14	0.99 (0.82 – 1.19)	0.88
Randomized to receive intervention (n = 8,054)	0.97 (0.73 – 1.30)	0.86	0.36 (0.18 – 0.71)	<0.01	1.32 (0.94 – 1.84)	0.11
Self-reported postmenopausal hormone therapy use ever							0.36
No (n = 17,252)	0.89 (0.74 – 1.08)	0.23	0.57 (0.37 – 0.89)	0.01	1.01 (0.81 – 1.25)	0.95
Yes (n = 18,731)	1.02 (0.82 – 1.27)	0.86	0.70 (0.44 – 1.13)	0.14	1.12 (0.88 – 1.44)	0.36

Abbreviations: HR, hazard ratio; CI, confidence interval; P, P value; P – interaction, P value for interaction with intervention

The ITT and per-protocol results were similar (Table 3). In the per-protocol analysis, CaD was not also associated with HF in the overall cohort nor in the high-risk subgroup but was associated with a significantly reduced risk of HF in the low-risk subgroup (Table 3). Similar CaD effects were observed among women who did not report taking personal vitamin D and calcium supplements and were also independent of competing events when estimated with the IPCW methods (Supplemental Table 3). Regardless of baseline risk status of HF, mean total calcium and vitamin D intake and serum 25(OH)D levels (available for only 2,012 women) were relatively lower in the incident HF cases compared to the non-cases (Supplemental Table 4).

Table 3.

Results of Intention-to-treat and Per-protocol Analysis Estimating the Association between CaD Supplementation and Heart Failure Incidence - The Vitamin D and calcium (CaD) Trial of the Women’s Health Initiative (WHI) Study, 1995 – 2005

Study population	Intention-to-treat analysis (N = 35,983)			*Per-protocol analysis (N = 23,601)
	Total	HF cases, n (rate/10,000 person-years)	HR (95% CI)	Total	HF cases, n (rate/10,000 person-years)	HR (95% CI)
Overall
Control	17,960	381 (29.8)	1.00	11,993	190 (21.9)	1.00
CaD	18,023	363 (28.2)	0.95 (0.82 – 1.09)	11,608	188 (22.4)	1.02 (0.84 – 1.25)
Low-risk
Control	9,227	96 (11.7)	1.00	6,320	51 (11.0)	1.00
CaD	9,307	61 (7.0)	0.63 (0.46 – 0.87)	6,186	30 (6.6)	0.60 (0.38 – 0.94)
High-risk
Control	8,733	285 (46.9)	1.0	5,673	139 (34.4)	1.00
CaD	8,716	302 (49.5)	1.06 (0.90 – 1.24)	5,422	158 (40.9)	1.19 (0.95 – 1.49)

Abbreviations: CaD, calcium plus vitamin D trial; HR, hazard ratio; CI, confidence interval; CI, confidence interval; P, P value

^*Analysis are based on data from women with at least 80% adherence to study protocol

Discussion

Vitamin D and calcium supplementation was associated with lower risk of HF in a subgroup of women without preexisting HF precursors at baseline but had no effect in those with these conditions. These findings were independent of baseline total vitamin D and calcium intake and persisted in per-protocol analysis. Results from a recent meta-analysis of 17 trials (n = 12,440), including the MRC RECORD trial, showed that vitamin D alone (vs placebo) was associated with a 21% lower risk of cardiac failure (HR, 0.79 [95% CI: 0.60 to 0.98]; P = 0.03) among men and women with a mean/median age > 60 years old²⁸. It is not clear whether the observed effect of vitamin D supplementation might have differed between baseline populations stratified by preexisting HF precursors in the meta-analysis population.

The basis for the differential effects of CaD intervention on HF incidence by baseline risk status is uncertain. Data from animal models suggest vitamin D may regulate cardiac functions, at least partially, via interaction with the VDR in cardiaomyocytes^{2, 3}. Through a receptor-mediated mechanism, 1,25(OH)₂D regulates intracellular calcium homeostasis and calcium ion uptake in ventricular cardiac muscle cells to modify cardiac contractility²⁹. An intermediate pathophysiological model involving the up-regulation of the renin-angiotensin-aldosterone system (RAAS) has been observed in both human and animal studies. RAAS plays a major role in HF pathogenesis by regulating blood pressure, cardiac contractility, electrolyte homeostasis, and eccentric hypertrophy of the myocardium^30,31. It has been demonstrated that a knock-out model of mice without the VDR develop high blood pressure, cardiac enlargement, and experience increased activation of the RAAS³². The activation of RAAS by low vitamin D status has recently been suggested in humans based on observational studies^{33, 34}. However, a meta-analysis of 10 trials did not show an association between vitamin D (alone or with calcium) supplementation and reduction in either systolic or diastolic blood pressures¹⁶.

Hyperparathyroidism has also been postulated as an intermediate process in the vitamin D-HF pathophysiology. Data from end-stage renal disease (ESRD) patients suggests secondary hyperparathyroidism and elevated parathyroid hormone (PTH) levels were associated with low vitamin D levels caused by failure of the kidneys to convert 25(OH)D to the metabolically active form of 1,25(OH)D₂^{2, 8–10}. Chronic exposure to PTH has been reported to be associated with poor myocardial structure and functioning as well as elevated blood pressure and accelerated atherosclerosis. It is therefore plausible the lower risk of HF observed in the low-risk subgroup is a result of the ability of the vitamin D supplements to prevent or mitigate the process of hyperparathyroidism.

We expected the high-risk subgroup would benefit more from the intervention, based on evidence that vitamin D deficiency is relatively higher in this subgroup in the general population, but our data did not confirm this hypothesis. One plausible explanation for the lack of an association between CaD and HF incidence in the high-risk subgroup is that HF pathogenesis was already too advanced to be influenced by the CaD supplements. Thus, in this high-risk population, any potential benefit related to HF prevention may have been diminished. Our data suggest that the high-risk group may indeed have poorer health status compared to those free of these comorbidities; baseline factors associated with poor health and risk of HF, such as older age, obesity, and hypercholesterolemia, were predominantly prevalent in this high-risk subgroup (Table 1).

It is also plausible that medical therapy for management of the cardiovascular risk factors in the high-risk group may have modified the effect of CaD intervention on HF. We therefore evaluated whether the use of CVD medications such as angiotensin concerting enzymes (ACE) inhibitors, angiotensin II receptor blockers (ARB), statins, beta-blockers, or calcium channel blockers were effect modifiers. Individually and collectively (composite of CVD medications), CVD medications did not significantly interact with CaD in relation to HF except ARB (P for interaction = 0.04) (Supplemental material). This supports the possibility that CaD supplementation may interact with medications used to treat comorbidities, thus making any direct effect of the intervention on HF outcome more difficult to detect. Additional studies on this topic are warranted.

This study contributes to the growing literature on the association between vitamin D and lower incidence of HF. This study is probably the first to demonstrate that CaD supplementation may be associated with lower risk of HF among postmenopausal women without preexisting hypertension, diabetes, CHD, or other CVD but not in those with these conditions. The study had adequate statistical power to perform the pre-specified subgroup analysis. The randomized trial design nature of the data minimized the effect of potential confounders.

Despite the strengths of this randomized trial, limitations of the study also warrant consideration. First, as a result of combining calcium with vitamin D as the intervention, it is unclear whether the observed associations are due to one or both of these dietary supplements. However, based on the meta-analysis²⁶ report that suggested an association between vitamin D (alone) supplementation and lower risk of HF and other reports that suggest calcium (alone) may be associated with increased risk of CVD events, it can be inferred that vitamin D was the active agent responsible for the prevention of HF in the WHI trial. Second, the CaD trial was limited to postmenopausal women, 50 – 79 years old, which may preclude generalizability of the findings to men or to younger women. Lastly, for ethical reasons, participants were allowed to continue consumption of both vitamin D and calcium supplements within the recommended dietary allowance guidelines put forth by the Institute of Medicine. However, consumption of personal vitamin D and calcium supplements did not modify the effect of CaD on HF incidence.

Our findings suggest that a very low cost daily supplementation with vitamin D (400 IU) plus calcium (1,000 mg) may be an effective primary prevention strategy for HF in postmenopausal women free of preexisting cardiovascular conditions. However, it appears to be of little value in preventing HF in the overall trial population and the subgroup of postmenopausal women who already have CVD conditions, including CHD, diabetes, and hypertension. These findings, if confirmed by other research, may have important public health and clinical implications.