Invited Commentary: The Association of Low Vitamin D With Cardiovascular Disease—Getting at the “Heart and Soul” of the Relationship

Abstract

Low concentrations of 25-hydroxyvitamin D have been consistently associated with cardiovascular disease (CVD) in many observational studies. In an analysis published in this issue of the American Journal of Epidemiology, Welles et al. (Am J Epidemiol. 2014;179(11):1279–1287) used data from 946 participants with stable CVD who were enrolled in the Heart and Soul Study (San Francisco Bay Area, 2000–2012) and found that the association of low 25-hydroxyvitamin D with increased secondary CVD event risk was attenuated after adjustment for parathyroid hormone level, suggesting that parathyroid hormone may mediate this association. They used observational data to gain insight into potential mechanisms underlying the association between vitamin D and CVD risk. Their study focused on secondary CVD events, whereas many previous observational studies have focused on incident CVD events among persons without a history of CVD. In this commentary, we place the study by Welles et al. in context with the existing literature and propose future directions for vitamin D research. We highlight a number of methodological concepts that are important in analyzing vitamin D data, including racial differences in vitamin D concentrations and adjustment for seasonal variation in vitamin D concentrations. We agree that randomized controlled trials should be conducted before making guidelines for screening and treating vitamin D deficiency for the prevention of CVD events.

Low concentrations of 25-hydroxyvitamin D have been consistently associated with cardiovascular disease (CVD) in observational studies (1). This association may be a result of reverse causation, whereby the lower levels of vitamin D seen in persons with CVD are a result of these persons’ being less healthy and less likely to go outdoors and be exposed to sunlight (2). Alternatively, this association may be mediated by a biological mechanism whereby low vitamin D concentrations cause increased risk of CVD events; this possibility is supported by the prospective nature of many observational studies in which vitamin D was measured long before the occurrence of CVD events.

In this issue of the American Journal of Epidemiology, Welles et al. (3) provide evidence from approximately 1,000 participants with stable CVD enrolled in the Heart and Soul Study (San Francisco Bay Area, 2000–2012) that the association between 25-hydroxyvitamin D and CVD risk may be mediated by parathyroid hormone (PTH) levels. The authors also note that this association was strongest in patients with diabetes, chronic kidney disease, and albuminuria (3). This study suggests the possibility that correction of vitamin D deficiency may reduce CVD risk, possibly via decreases in PTH concentrations. The authors call for randomized controlled trials to determine whether vitamin D supplementation does decrease CVD risk.

Vitamin D and PTH are tightly inversely correlated; PTH concentration rises with concurrent vitamin D deficiency, and treating vitamin D deficiency results in lower PTH levels. A recent review suggested that higher levels of PTH are associated with increased risk of CVD (4). However, because of the tight inverse relationship between vitamin D and PTH, it is difficult to determine directionality in the association of these biomarkers with CVD risk. Although PTH level may be one mechanism through which low vitamin D concentrations lead to increased CVD risk via increased vascular remodeling (5), there are many other plausible pathways. Other hypothesized mechanisms include activation of the renin-angiotensin-aldosterone system (6), elevation of blood pressure (7), an adverse glucose/metabolic profile (7), increased inflammation, and increased atherogenesis (8).

The results of this study must be considered in the context of the existing observational literature on the association of low vitamin D concentrations with CVD risk (1). As Welles et al. note (3), many prior observational studies did not adjust for PTH levels. However, 2 larger studies, the Ludwigshafen Risk and Cardiovascular Health (LURIC) Study (n = 3,258) (9) and the Multi-Ethnic Study of Atherosclerosis (MESA) (n = 6,436) (10), did adjust for PTH and found that low concentrations of 25-hydroxyvitamin D remained significantly associated with increased CVD risk. It is important to note that the study by Welles et al. (3) and the LURIC Study (9) examined the association between vitamin D and CVD events among persons with established CVD (secondary events), whereas MESA (10) examined this association among persons with no history of CVD (primary events).

Given the numerous observational studies showing the association of CVD risk with vitamin D deficiency (1), it is time to focus on randomized controlled trials to determine whether vitamin D supplementation reduces CVD risk, as Welles et al. state (3). In these trials, an important distinction will be between primary and secondary prevention of CVD. It is possible that vitamin D supplementation may have more relevance in the primary prevention of atherosclerosis (e.g., the ongoing Vitamin D and Omega-3 Trial (VITAL) (11)) and less relevance for secondary prevention among persons with established CVD, where the incremental benefit of adding vitamin D supplementation may be much smaller in the context of other medical therapies that significantly reduce the risk of secondary CVD events. The possible importance of vitamin D supplementation in primary prevention is supported by in vivo studies suggesting a role of vitamin D in suppressing macrophage cholesterol uptake and the transition to foam cells (12).

Although there are a number of important randomized clinical trials of vitamin D supplementation that are currently ongoing, there are some challenges in their design that should be taken into consideration. First, it is difficult to ever have a true placebo arm in these trials, since people can get vitamin D through diet and exposure to sunlight. Second, many trials do not use vitamin D deficiency as an enrollment criterion. Vitamin D supplements may only benefit persons with vitamin D deficiency and may have no benefit in people with adequate levels of vitamin D. Given the diversity of enrollment levels of vitamin D, the effect of a standard dose of vitamin D (e.g., 2,000 IU/day in the VITAL study (11)) probably will not be the same in all participants. For participants with a very low starting 25-hydroxyvitamin D level, this dose may not be enough to attain an adequate level of vitamin D, and for participants with an adequate starting level of vitamin D, this dose may not be necessary or may be harmful because of the increased risk of kidney stones (13). However, there has been much controversy about what defines adequate levels of 25-hydroxyvitamin D. Some studies have suggested a U-shaped association of vitamin D with CVD risk (14, 15), whereas others, including the study by Welles et al. (3), have suggested a threshold effect at 20 ng/mL. A clinical trial of the effect of attaining a particular serum level of 25-hydroxyvitamin D (e.g., >20 ng/mL vs. >30 ng/mL) or a certain level of bioavailable vitamin D may be more clinically relevant than an analysis of fixed doses of vitamin D supplements.

One issue that is important to consider in the interpretation of Welles et al.'s results (3) is racial differences in 25-hydroxyvitamin D levels. Racial differences in vitamin D levels, where blacks have lower levels than whites, have been shown in many studies (16). Further, there is evidence that associations of 25-hydroxyvitamin D with CVD or stroke may differ by race, with increased risk seen in whites but not in blacks (10, 17). The study by Welles et al. was comprised of approximately 16% blacks and 61% whites, and the authors did not find any significant racial differences in the association of low vitamin D with risk of a second CVD event (3). However, results were not presented separately by race, nor were race-specific cutpoints used. Although blacks have been shown to have lower levels of total 25-hydroxyvitamin D than whites (16), blacks also have lower levels of vitamin D-binding protein, which results in estimated concentrations of bioavailable vitamin D comparable to those of whites—suggesting that levels of bioavailable vitamin D may be the most clinically relevant measurement to study (18).

The vitamin D concentrations in the Heart and Soul Study were measured using the gold standard method of mass spectrometry. This method avoids the problems with assay drift that occurred in many prior studies that used radioimmunoassay to measure vitamin D (19). Vitamin D concentration varies with season (20), and many analytical approaches have been suggested and used to account for this variation (21). Welles et al. adjusted for date of sample collection in their regression models (3). However, a study by Wang et al. (21) suggested that adjustment for date of sample collection sometimes results in a nonconservative bias. Wang et al. recommended using a month- or season-specific cutpoint approach to reduce the possibility of biasing results away from the null (21). More recently, a cosinor approach has been used to account for seasonal variation in vitamin D concentrations (10, 20). This body of literature supports the notion that the use of analytical methods (e.g., month-/season-specific cutpoints, cosinor approach) to account for seasonal variation in vitamin D concentration may help to reduce nonconservative biases and more accurately inform clinical decisions.

In conclusion, the new findings by Welles et al. (3) support other observational literature on the association between vitamin D deficiency and increased CVD risk and add to the literature by suggesting possible mediation of this association by PTH. At this point, there are a number of observational studies supporting this relationship (1), but data from randomized clinical trials, such the VITAL study (11), are still being collected. To get at the “heart and soul” of the relationship, researchers in future studies will need to be careful about methodological issues accounting for seasonal change, to explore racial differences in the associations seen, and to determine whether the association with low vitamin D is causal or confounded by poorer health status. There is a need for randomized controlled trials to be completed before guidelines are made for screening and treating vitamin D deficiency for the prevention of CVD events.