Professor Ian Reid, Faculty of Medical and Health Sciences, University of Auckland, Private Bag 92019, Auckland, New Zealand. Tel.: (+649) 3737 599 extn 86259; Fax: (+649) 308 2308. E-mail: firstname.lastname@example.org
Calcium supplementation is widely used for the prevention of osteoporosis in postmenopausal women and in men. While there has been ongoing debate regarding its effectiveness in fracture prevention, the underlying assumption has been that, even if it was not particularly effective, at least it was safe. The recent finding of the Auckland Calcium Study that myocardial infarctions were more common in women randomised to calcium calls this assumption into question, and consideration of vascular event data from other calcium trials does not refute the Auckland findings. Meta-analyses of these data will be necessary to settle this matter. It is already accepted that calcium supplements increase vascular risk in patients with renal compromise, even in those not yet requiring dialysis. Also, there is substantial epidemiological evidence that serum calcium levels in the upper part of the normal range are a risk factor for vascular disease, and that calcium supplements acutely elevate serum calcium – a combination of findings that lends plausibility to supplementation increasing vascular risk. As there are reasonable grounds for doubting the safety of calcium supplements, and as the evidence for their efficacy in fracture prevention remains marginal, we suggest that there should be a reappraisal of their role in the management of osteoporosis, with a greater emphasis on agents known to prevent fractures.
Calcium supplementation is widely recommended to postmenopausal women for prevention of osteoporosis. As osteoporosis awareness increases in men, supplement use in this group is also increasing. While calcium supplements clearly slow bone loss in both sexes,1,2 their effects on fracture risk are less certain.3,4 However, other benefits from the use of calcium supplements have been suggested, including effects on colon cancer, body weight and cardiovascular disease.5 There is evidence that calcium intake might impact on serum lipid concentrations from observational clinical studies and from randomised trials which have shown beneficial changes in serum cholesterol and/or its subfractions.6–12 These changes might result from binding of calcium to fatty acids and bile acids in the gut, leading to reduced fat absorption7,13,14 and could also contribute to the inverse association between calcium intake and body weight found in several observational studies.15–17 There is also evidence that calcium supplementation reduces blood pressure,18,19 and that cardiovascular mortality is lower in hard water areas20 and in individuals with high calcium intakes.21–24 These findings all suggest that calcium use might impact favourably on cardiovascular disease. If this were the case, the vascular effects of calcium supplementation could have a greater influence on total morbidity and mortality than its effects on bone because of the prevalence of vascular disease and the importance of its consequences.
While the potential positive benefits of calcium supplements on cardiovascular health have been promoted by some, it has not had universal support. There has been a growing awareness that calcium intake, vascular calcification and mortality are linked in patients with renal failure25 and some have suggested that such adverse effects are seen not only in uraemia.26 As one vascular biologist asked: ‘if calcium is accumulating in the arteries while leaking from the skeleton, can we assume that supplemental calcium goes selectively to bone?’27 To address the uncertainty surrounding the vascular effects of calcium supplementation, a topic of considerable importance to the health of postmenopausal women, we prespecified a number of vascular end-points in our recent randomised controlled trial of calcium supplementation in older women.2,28
The Auckland Calcium Study
This was a randomised controlled trial of calcium supplementation in normal postmenopausal women, designed primarily to assess the effects of calcium on bone density and fracture incidence over a period of 5 years. A total of 1471 healthy older women were randomised to calcium (1 g daily, as the citrate) or placebo and followed for 5 years. Women receiving therapy for osteoporosis or taking calcium supplements were ineligible, as were those with any other major ongoing disease or with serum 25-hydroxyvitamin D < 25 nmol/l. Complete follow-up was achieved in 90% of subjects.
During the study, 34 myocardial infarctions, 57 strokes and 6 sudden deaths were reported and verified. The relative risks of these events in the calcium group were myocardial infarction 2·12 (1·01, 4·47), stroke 1·42 (0·83, 2·43), sudden death 1·01 (0·20, 4·99). Each of these events was adjudicated by physicians, and a national database of hospital admissions was also searched for any events not reported by study participants. The Kaplan–Meier plot for myocardial infarction showed that the groups progressively diverged from about 2 years (Fig. 1). The rate ratio for composite end-point was 1·43 (P = 0·043). Based on the event rates in this study, the number of women needed to treat for 5 years to cause one myocardial infarction was 44, to cause one stroke was 56 and to cause one cardiovascular event was 29. By comparison, the number needed to treat to prevent one symptomatic fracture was 50.
Other studies assessing the effects of calcium on cardiovascular disease
These surprising findings from the Auckland Calcium Study provide a stimulus to assess the effects of calcium on vascular disease events in other similar trials.
Trials of calcium monotherapy
To this end, we searched MEDLINE, EMBASE, and Cochrane CENTRAL (1966–November 2007), reference lists of meta-analyses of calcium supplementation and two clinical trial registries for randomised, placebo-controlled trials of calcium supplementation (≥500 mg/day) with ≥100 participants and duration >1 year. Many studies of calcium supplementation make no mention of vascular events. 29–31 Those which do are as follows. The RECORD study, a secondary fracture prevention study of participants of similar age to those in the Auckland study, reported a trend to higher death rates in those allocated to calcium compared with placebo (18·5%vs 16·3%). Prince32, in a study of 1460 postmenopausal women (mean age 75 years) randomised to calcium carbonate or placebo over a 5-year period, found the hazard ratio for the diagnosis of ‘incident ischaemic heart disease’ was 1·12 (95%CI 0·77–1·64). The findings of these studies are congruent with those from the Auckland study. Recently, we reported a study of calcium supplementation in 323 normal men randomised to two doses of calcium or to placebo and followed for 2 years.1 Despite its low power to address vascular event end-points, it is noteworthy that all of the coronary heart disease events (9 in 5 men) occurred in men randomised to calcium. In contrast, a study of 1200 women over 4 years did not suggest an adverse effect of calcium supplementation on vascular risk,33 although use of nontrial calcium and oestrogen supplements was permitted in this study. Taken together, these randomised trials do not allay the concerns regarding the cardiovascular safety of calcium supplementation, raised by the Auckland Calcium Study.
Women’s Health Initiative
The Women’s Health Initiative (WHI) is by far the largest study of calcium supplementation, but calcium was administered together with vitamin D in this study. A total of 36 000 women were randomised to calcium 1 g/day (as the carbonate) and vitamin D (400 IU/day), and followed over 7 years.34 While the primary end-point was fracture, the vascular outcomes have also been described35 and some of these are presented in Table 1. The hazard ratio for myocardial infarction or coronary heart disease death in the whole cohort was 1·04 (95%CI 0·92–1·18), but for those without risk factors for coronary heart disease it was 1·19 (95%CI 0·97–1·59). For the composite end-point of myocardial infarction, coronary heart disease death, coronary artery bypass graft or percutaneous coronary intervention, the hazard ratio in the entire cohort suggested an adverse effect (1·08, 95%CI 0·99–1·19). Vascular risk associated with randomisation to calcium/vitamin D was related to body mass index (BMI) (P= 0·04) such that women with higher BMI were at lower risk with active calcium/vitamin D supplementation. For participants with BMI of 25–30, the hazard ratio for coronary heart disease was 1·18, and for BMI < 25 it was 1·16. It would be of interest to know what the hazard ratios are for the composite end-point (including percutaneous coronary intervention) in the nonobese population, because this is the group most at risk of osteoporosis and so most likely to use calcium supplements.
Table 1. Risks of CHD Associated with Ca/D by Subgroups in WHI
None of the hazard ratios is statistically significant.
CHD, coronary heart disease; BMI, body mass index; WHI, Women’s Health Initiative.
CHD risk factors
The subjects in the WHI differed in a number of respects from those in the Auckland Calcium Study – they were 12 years younger, heavier (mean BMI 29 vs 26), had higher calcium intakes as a result of use of nontrial calcium supplements (mean 1150 mg/day), and 50% were taking hormone replacement therapy. In the WHI, hazard ratios for vascular events in the calcium/vitamin D-supplemented group tended to be higher in older subjects, in those with lower calcium intakes, and in those with a lower BMI. If the tertiles of each of these variables that most closely match the characteristics of the subjects in our own study are chosen, then the calculated relative risk of coronary heart disease associated with calcium use in WHI is 1·36, very similar to the Auckland study. The use of a less bioavailable supplement in the WHI may have reduced the size of any adverse effect, as may the poor compliance with study medication. The co-administration of vitamin D may also have influenced the outcome, because there is evidence that calciferol alone reduces all-causes mortality.36 The use of oestrogen and personal calcium supplements might also have obscured any vascular effects of the calcium supplementation.
Following the publication of the vascular event data from the Auckland Calcium Study, the investigators in the Kuopio Osteoporosis Study turned to their large observational database to determine whether a similar effect was evident there. In a cohort of 10 555 women born between 1932 and 1941, who were free of coronary heart disease in 1994, they noted the use of calcium (with or without vitamin D) in 26%. After a mean period of follow-up of 6·6 years, there were 513 new cases of coronary heart disease, the hazard ratio in those using calcium being 1·24 (95% CI 1·02–1·52).37 While limited by the observational study design, this result is consistent with the randomised trial data cited earlier. In contrast, a large study of calcium supplement use in American men did not find a significant effect on the risk of ischaemic heart disease.38
Calcium effects in renal failure
There is also a literature in patients with renal failure which assesses the effects of calcium supplementation on vascular outcomes. Comparisons of calcium supplements (used as phosphate binders) with sevelamer (a noncalcium, phosphate-binding resin) show that calcium users have more rapid aortic and coronary artery calcification,39–41 and higher all-cause mortality.42 In young dialysis patients, coronary artery calcification is more common in those with higher calcium intakes,43 and there is a positive relationship between serum calcium and cardiovascular events in patients on dialysis.44 Perhaps more relevant to the population at risk of osteoporosis, calcium supplementation has also been shown to accelerate vascular calcification in predialysis patients with a mean glomerular filtration rate (GFR) of approximately 30 (SD 20) ml/min45. As GFR in the Auckland Calcium Study was 61 (SD 11) ml/min, there is considerable overlap between these two groups in terms of renal function, suggesting that these changes might occur in the ‘normal’ elderly population as well. If renal function were to interact with the vascular effects of calcium, then vascular toxicity is less likely to be an issue in younger postmenopausal women.
How might calcium supplementation cause heart disease?
If we accept that there is a body of evidence suggesting that calcium supplementation has an adverse effect on vascular health, then it becomes important to consider the potential mechanisms underlying this. Of possible relevance is another group of observational studies relating serum calcium concentrations to vascular health.
Serum calcium concentrations and vascular disease
Rubin et al.46, in a community-based study of older men and women, have reported that carotid artery plaque thickness is increased by one-third in subjects in the highest quintile of serum calcium concentrations, in comparison with those in the lowest quintile. We have recently quantified aortic calcification in the baseline lateral spine DXA images from women in the Auckland Calcium Study and found that for each increase in total serum calcium of 0·1 mmol/l, the odds ratio for the presence of aortic calcification is 1·23 (95% CI 1·08–1·40), after adjustment for multiple other factors, including age, weight, smoking and history of heart disease (M Bolland et al., in preparation).
In the Tromso study, it has been found that serum calcium is a highly significant predictor of myocardial infarction in men (odds ratio 1·2 per 0·1 mmol/l increase in serum calcium), and a similar but a nonsignificant trend was seen in women.47 In another Swedish study, 2183 men were followed for 18 years and serum calcium was again found to predict myocardial infarction, with the normal range of serum calcium levels accounting for a 2·5-fold range in risk.48 In an American study of 15 732 men and women followed over 13 years, adjusted serum calcium at baseline was significantly related to risk of myocardial infarction (hazard ratio/SD 1·05 [1·01,1·10]), stroke (HR 1·37 [1·28-1·46]), and death (HR 1·21 [1·16,1·21]).49 Unpublished analyses from the MORE Study indicate that incident cardiovascular events increase 17% for each standard deviation increase in baseline serum calcium (Slinin et al., personal communication). This suggests that the changes in the blood vessel walls, demonstrated in the Rubin and Bolland studies, translate into an increased risk of vascular events, though we must be cautious in inferring causation from such observational data.
Population data from men in Malmo demonstrate that there is an increased mortality associated with higher calcium levels. Leifsson50 assessed mortality for men under the age of 50 in relation to their serum calcium concentration, in a population-based study of 21 131 Swedish men under the age of 50. The death rate for those with serum calcium levels between 2·45 and 2·5 mmol/l was 1·3 (95% CI 1·1–1·4) times higher than that in men with baseline calcium between 2·3 and 2·45 mmol/l. For those with serum calcium between 2·5 and 2·55 mmol/L, the odds ratio of death was 1·5 (95% CI 1·3–1·8). The excess mortality appeared to be predominantly cardiovascular. These findings are consistent with evidence of increased mortality in primary hyperparathyroidism, in which condition serum calcium levels are often near the upper end of the normal range.51
Effect of calcium supplementation on serum calcium
These findings pose the question of what effect calcium supplementation has on serum calcium. We have previously shown that ionised calcium concentration increased from a mean of 1·22–1·30 mmol/l following supplementation with 1 g of calcium, either as the citrate or as the lactate-gluconate (Fig. 2).52 With the less-soluble carbonate salt, the increase was 0·05 mmol/l, and the same dose of calcium as bone meal only increased ionised calcium by 0·03 mmol/l. Following ingestion of the two most soluble calcium preparations, all individuals had ionised calcium concentrations above the upper end of the laboratory reference range at 3 h. Thus, the changes in serum calcium following ingestion of a calcium supplement are greater than the serum calcium differences which have been shown in observational studies to be associated with vascular pathology, cardiovascular events and cardiovascular mortality, though whether such large changes are maintained with chronic calcium dosing is unknown.
These changes are in marked contrast to the effects of dairy products on serum calcium. For instance, subjects taking 1200 mg of calcium as fortified skim milk show increases in total serum calcium of only 0·03 mmol/l at 3 h,53 (Fig. 3) in contrast to the changes in ionised calcium shown in Fig. 2. Further, this study found that skim milk fortified with calcium carbonate caused larger increases in serum calcium than did product fortified with ‘milk calcium’. This suggests that the increase in ionised calcium following comparable doses of a dairy product are only one-sixth of those seen following the ingestion of a soluble calcium supplement. These data might have important implications for the comparative vascular safety of calcium salts and food sources of calcium, and potentially point the way to how calcium intake might be maintained for bone health without adversely affecting vascular health.
How might serum calcium impact on heart disease?
If serum calcium impacts on vascular disease, then how is this likely to be mediated? Vascular calcification is no longer thought to be the result of a simple physico-chemical precipitation of calcium and phosphate. Rather, it is a regulated process involving a number of inhibitors of crystal formation, including fetuin A, matrix Gla protein, pyrophosphate, osteoprotegerin and bone morphogenic protein-7. We have looked for changes in matrix Gla protein and osteoprotegerin in 80 women treated with calcium or placebo over a period of 5 years. As shown in Fig. 4, there is no suggestion that long-term calcium use influences these regulators of mineralisation. This does not rule out more transient changes in these factors following calcium ingestion. For instance, it has recently been demonstrated that phosphate concentrations acutely regulate matrix Gla protein secretion from chondrocytes.54 As calcium ingestion results in transient reductions in phosphate, this is a potential mechanism by which calcium use could directly impact on tissue calcification. Perhaps the most likely of these calcification modulators to be influenced by calcium ingestion is pyrophosphate. Pyrophosphate is present in extracellular fluid in micromolar concentrations and is in a dynamic equilibrium with calcium, with which it forms complexes. When ionised calcium concentrations increase following calcium ingestion, there would be expected to be a reciprocal decline in pyrophosphate, with associated loss of inhibition of mineralisation.
A second mechanism by which serum calcium might influence vascular calcification is via its binding to the calcium-sensing receptor, because these receptors are expressed on vascular smooth muscle cells. Exposure of cultures of these cells to increased concentrations of calcium or to gadolinium results in down-regulation of the calcium-sensing receptor and to increased mineralisation of the cultures.55 A role for this receptor is also suggested by the study of Marz56 in which a polymorphism of the calcium-sensing receptor gene (the S-allele, in which an alanine to serine substitution takes place in codon 986) was found to be associated with a 22% increase in prevalence of coronary artery disease (95% CI 1·02–1·47) for each S-allele present, and a 30% increase in myocardial infarction (95% CI 6–60%) per S-allele. Cardiovascular mortality was increased 38% (95% CI 13–67) per S-allele. The presence of the S-allele is also associated with increased serum calcium of approximately 0·02 mmol/l but adjustment for this did not remove the risk associated with the S-allele. The latter finding could result from the lability of serum calcium over time, and the failure of a single measurement to represent circulating plasma levels over the decades during which atherosclerotic disease will have developed. An alternative explanation could be that circulating calcium acts through the calcium-sensing receptors in vascular tissue to influence development of atherosclerosis, the S-allele altering tissue responsiveness to circulating calcium. This possibility is supported by the finding that calcium-sensing receptor agonists promote mobilisation of soft tissue calcification in uraemic rats, apparently by activating macrophages.57 Calcium-sensing receptors are also found on platelets,58 so calcium concentrations might impact on function of these cells which are also implicated in the pathogenesis of vascular events.
A possible pathogenic pathway that fits together the various factors discussed earlier is shown in Fig. 5.
Where to now?
The information reviewed earlier raises important questions relating to public health and clinical practice. Because calcium is a key constituent of all diets, the possibility that it might influence vascular disease should be resolved as quickly as possible. The most straightforward solution to this problem would appear to be the undertaking of a large definitive trial. A possible design would be similar to the Women’s Health Initiative calcium study, but using calcium alone, only recruiting subjects not self-administering calcium supplements, and using a calcium salt with high bioavailability. Based on our own trial data, a placebo-controlled trial of 13 500 women over 5 years would have 80% power to detect a 20% difference in the composite end-point of myocardial infarction, stroke or sudden death. Such a study may well address the remaining clinical uncertainties, but would be both unethical and impractical, because the primary hypothesis is one of harm. It is analogous to suggesting that remaining uncertainties regarding the safety of cigarette smoking should also be addressed with a randomised, controlled trial. Therefore, the only way to increase our knowledge in this area is to meta-analyse the data that have already been collected.
We are undertaking such an analysis, following standard guidelines, and have identified 15 studies of calcium monotherapy which are of sufficient size and duration to be able to contribute significant numbers of events to this analysis. Together these studies comprise almost 13 000 subjects studied over a mean period of 4 years. Through collaborations with investigators of some of these studies, we will have individual data for over 8000 subjects together with trial level data for another 4000 subjects, giving us some information on cardiovascular events for 93% of the total possible cohort.
Until the meta-analysis results are available, both doctors and patients need to find a practical way forward. We believe that the available data raise the significant possibility that calcium supplementation results in a modest increase in cardiovascular risk. This needs to be balanced against its potential benefit in the prevention of fractures. While there is unequivocal evidence of beneficial effects of calcium on bone density, meta-analyses of its effects on fracture show effects which are of borderline significance, both statistically and clinically (approximately 10% risk reduction in all fractures).3 Therefore, we would suggest that the widespread use of an agent which may have vascular risks comparable to those of hormone replacement therapy, but is considerably less effective in preventing fractures is unwise, and that those at significant risk of fracture should be encouraged to use agents (such as bisphosphonates) with established anti-fracture efficacy. If calcium is to be used, we would advise moderation, aiming for a total calcium intake (diet plus supplement) of about 1 g/day. We would also suggest that food sources of calcium may be preferable to pharmacological supplements because they cause much smaller perturbations of serum calcium. In subjects unable to tolerate dairy products or other food sources of calcium, the use of small doses of insoluble calcium salts may be justifiable, but it is probably sensible to give these in divided doses with meals so that their rate of absorption is minimised.
As food calcium and that from pure salts have different acute effects on serum calcium, it is important that data generated in studies of calcium supplements are not automatically applied to calcium-rich foods. A critical item on the research agenda is to compare the effects of these two calcium sources on markers and regulators of vascular health. While these clinical studies are being undertaken, laboratory studies of the impact of calcium on vascular biology are vitally important. This is already an important area of enquiry in the context of renal failure, but the focus must be broadened to include the biology of vascular calcification and atherosclerosis in subjects with normal renal function.
We are grateful to Biomedica, Vienna, for donation of the kits used in producing the data shown in Fig. 4. Supported by the Health Research Council of New Zealand.