Dietary Calcium and Blood Pressure

Methods

We attempted to identify, review, and include all randomized clinical trials assessing the effect of dietary calcium supplementation on blood pressure. We identified trials through a search of previous review articles, a MEDLINE search (January 1982 to December 1993), a manual review of articles in major journals, and a review of abstracts from scientific meetings. We sought to identify all relevant studies that had been published through December 1993.

Independent reviewers abstracted data in duplicate from the identified studies. These data included, to the extent provided within each paper, mean systolic and diastolic blood pressure (at baseline and follow-up or the change between baseline and follow-up blood pressures), standard deviation or standard error for blood pressures, t-statistics or P values, calcium dose and formulation, duration of intervention, type of design (crossover or parallel), mean age of participants, percentage of participants who were male, percentage of participants who were white, initial blood pressure status, and whether the study was double-blinded. Measures of compliance were not reported for most studies and thus were not abstracted. For parallel trials, the blood pressures abstracted were the changes from baseline (final value − baseline value) for the intervention group minus the change from baseline for the control group. For crossover trials, the abstracted values were the final blood pressures for the intervention group minus those for the control group. When we abstracted studies that provided stratified data and analyses (by blood pressure status, race, sex, or calcium dose or formulation), we maintained the stratification so that we could pool the data by strata. A third author reviewed the duplicate abstraction forms for inconsistencies; discrepancies were reconciled by consensus.

We pooled data using weighted averages, with weights equal to the inverse of the variance of the observed effect (Appendix) [35]. Trials were excluded from pooling if the calcium supplementation was part of a multifactorial trial or if the data presentation posed analytic problems (see below). We also did separate analyses for trials in normotensive and hypertensive persons and for parallel and crossover design trials. Not all studies reported the variance measures needed to calculate weights for the pooled analysis. For these studies, we used a weighted average of the reported variances.

We used two methods to determine whether other factors, such as age, sex, calcium dose, and trial duration, modified the effects of calcium on blood pressure. First, in the nonparametric approach, we calculated Spearman correlation coefficients, which weight each trial equally. Second, because different trials were based on different numbers of participants, we did separate parametric weighted (by sample size) linear regression analyses of the estimate of the effect on each of the four factors.

Results

We identified 26 randomized clinical trials that assessed the effect of calcium supplementation on blood pressure. Two of these trials had two active treatment arms: One used two different calcium doses [25] and 1 used two different calcium formulations [17]. One trial stratified data by race [15] and another, by sex [9]. This stratification induces correlation between two pairs of effect estimates, whereas our statistical methods for pooling data require independence; however, the error is negligible (< 0.05 mm Hg). Because 3 trials reported data separately for hypertensive and normotensive persons, we abstracted 33 separate strata from the 26 trials, as shown in Table 1 and Table 2.

Trials Involving Normotensive Persons

Thirteen trials studied normotensive persons; 3 of these also included hypertensive persons. Mean participant age ranged from 21 to 48 years. Six trials included both sexes, 3 studied men only, and 4 studied women only; 4 included black persons. Nine trials were double-blinded, 3 trials were open, and 1 trial (by Vinson and colleagues [17]) was doubleblinded in one of the two methods used for calcium supplementation (gluconate and yeast). Eight trials had parallel designs, with sample sizes ranging from 14 to 471 persons; 5 trials had crossover designs, with sample sizes of 7, 8, 29, 32, and 50 persons, respectively. Duration of intervention and follow-up ranged from 1 week to 4 years (median, 8 weeks). Calcium doses in these studies ranged from 0.5 to 2.0 g. Except for two trials that used diets rich in calcium [18, 19] and both arms of Vinson and colleagues' trial [17], most trials used tablets containing calcium carbonate.

Systolic and diastolic blood pressure significantly decreased in 2 of these 13 trials. Seven trials had negative results consisting of either a nonsignificant beneficial effect or direct association between calcium intake and blood pressure; 4 trials had mixed results. Of the studies with mixed results, 3 showed a significant effect for either systolic or diastolic blood pressure, but not for both; in Vinson and colleagues' study [17], only the calcium yeast supplement affected systolic blood pressure.

Trials Involving Hypertensive Persons

Sixteen trials randomly assigned hypertensive persons, including the three trials that included both normotensive and hypertensive persons. The age range in most trials was broad, with most participants in middle age. All but one trial included both sexes to varying degrees, and Johnson and colleagues [11] studied only women. Five studies included black persons. All trials were double-blinded. Eleven studies used a crossover study design, with sample size ranging from 8 to 48 persons; five studies used a parallel design, with sample size ranging from 27 to 90 persons. Interventions lasted from 5 days to 4 years (median, 8 weeks). The calcium doses ranged from 0.4 to 2.16 g (median, 1.0 g).

Thirteen of the studies reported no significant reductions in blood pressure; among these, Johnson and coworkers [11] reported a decrease in systolic blood pressure of 20 mm Hg after 4 years of follow-up (P = 0.06), and Grobbee and Hofman [24] reported a decrease of 2.4 mm Hg in diastolic blood pressure (P = 0.11). Two trials had mixed results: McCarron and Morris [12] reported a decrease of 3.8 mm Hg in systolic blood pressure after 8 weeks (P = 0.02), and Lasaridis and associates [29] found a decrease of 5.4 mm Hg in systolic blood pressure (P = 0.05). In Saito and coworkers' study [32], systolic blood pressure decreased by 10.4 mm Hg and diastolic blood pressure decreased by 2.3 mm Hg, a significant effect (P = 0.01) for mean blood pressure.

Pooling of Trials

To quantify the effect of calcium supplementation on blood pressure, we pooled data from all trials, including those that studied both normotensive and hypertensive persons. We did not include four of the trials in the final pooling of data. Lyle and colleagues [15] did not provide unadjusted blood pressure data but rather adjusted the data for compliance with the intervention and for baseline variables such as age, skinfold thicknesses, and dietary calcium intake. We excluded the study by Van Beresteijn and associates [19] because it used a multifactorial intervention that included both magnesium and calcium. Bloomfield and coworkers' study [22] was excluded because it did not provide an effect estimate but rather reported that calcium had no effect. We excluded Tanji and colleagues' study [33] because the data could not be pooled: The trial design was crossover, but results were presented as if the study were parallel; in addition, the sample size at the end of follow-up was unclear because participants dropped out of the study at unspecified times. We thus pooled 28 arms or strata from 22 trials, yielding a total pooled sample of 1231 persons.

The weighted average changes in blood pressure were −0.89 mm Hg for systolic blood pressure (95% CI, −1.74 to −0.05 mm Hg) and −0.18 mm Hg for diastolic blood pressure (CI, −0.75 to 0.40 mm Hg) (Figure 1 and Figure 2). Because the Trials of Hypertension Prevention study [20] had the largest sample size and negative results, we separately analyzed the pooled data after excluding this study; this separate analysis indicated that the systolic blood pressure effect was −1.22 mm Hg (CI, −2.34 to −0.10 mm Hg) and that the diastolic blood pressure effect was −0.42 mm Hg (CI, −1.16 to 0.32 mm Hg).

We separately analyzed the pooled data for trials of normotensive and hypertensive persons and for parallel and crossover trials (Figure 1 and Figure 2). Pooled results for normotensive persons showed a systolic blood pressure change of −0.53 mm Hg (CI, −1.56 to 0.49 mm Hg) and a diastolic blood pressure change of −0.28 mm Hg (CI, −0.99 to 0.42 mm Hg); in hypertensive persons, the changes were −1.68 mm Hg for systolic blood pressure (CI, −3.18 to −0.18 mm Hg) and 0.02 mm Hg for diastolic blood pressure (CI, −0.96 to 1.00 mm Hg). For parallel trials, the change in systolic blood pressure was −0.77 mm Hg (CI, −1.87 to 0.34 mm Hg); for crossover trials, the change was −1.08 mm Hg (CI, −2.39 to 0.23 mm Hg). For parallel trials, the change in diastolic blood pressure was −0.19 mm Hg (CI, −0.95 to 0.58 mm Hg); for crossover trials, the change was −0.17 mm Hg (CI, −1.00 to 0.70 mm Hg).

We calculated Spearman correlation coefficients to determine whether age, sex, calcium dose, or trial duration was associated with calcium-mediated changes in blood pressure. The correlation between the effect on systolic blood pressure and mean age was −0.42 (P = 0.04), indicating a larger decrease among older persons; the correlation between systolic blood pressure and the percentage of men studied was 0.49 (P = 0.01), indicating a larger effect of calcium in women. No other correlations were significant. Because Spearman correlation coefficients apply equal weights to all trials, weighted linear regression analyses of the estimate of the effect on each of the four factors were done separately. All estimated slopes had P values greater than 0.10. We concluded that there were no consistent associations between the effect of calcium on blood pressure and age, sex, calcium dose, or trial duration.

Discussion

The pooled results of these trials of calcium supplementation show that this supplementation had only a small and inconsistent effect on blood pressure. These results indicate that a median calcium supplementation of 1 g/d is associated with a statistically significant decrease in systolic blood pressure of 1 to 2 mm Hg, both in the overall sample and among hypertensive persons; however, the effect is not statistically significant among normotensive persons or for diastolic blood pressure.

Publication bias must be considered when several studies are pooled to obtain an overall estimate. Trials with positive results are more likely to be published than trials with negative results. The omission of negative trials would show a stronger association between calcium and blood pressure than would be seen with their inclusion. Because we report only a small effect for hypertensive persons and a nonsignificant effect for normotensive persons, publication bias probably played an unimportant role in our pooled analysis.

After completing our search, we became aware of trials of calcium supplementation that focused on conditions other than hypertension. We therefore attempted to find data on blood pressure from trials of changes in bone mineral density. Although we could not review all of these studies, we identified more than a dozen studies through a bibliography prepared for the National Institutes of Health Consensus Development Conference on Optimal Calcium Intake. None of these reports provided data on blood pressure.

Three trials were excluded from the pooled analysis: Tanji and colleagues [33] found nonsignificant associations in a small sample of hypertensive persons (n = 19), whereas Lyle and colleagues [15] and Van Beresteijn and coworkers [19] found significant inverse associations in samples of normotensive persons (n = 75 and n = 53, respectively). Although these trials were not appropriate to include in the pooled analysis, they do provide some additional indication of a beneficial effect of increased dietary calcium intake.

These trials suggest that increased dietary calcium intake may slightly decrease blood pressure and, among other potential benefits, may help prevent osteoporosis. A National Institutes of Health Consensus Development Conference on Optimal Calcium Intake was held in June 1994 to address some of these issues. In the consensus statement [36], inconclusive findings regarding preeclampsia and colon cancer were reported (a large study of calcium intake and preeclampsia [37] is under way, with results expected in 1996). This conference also addressed the adverse effects of increased calcium intake. Although persons with a history of kidney stones may be at risk for stone recurrence with increased calcium intake, most persons have no adverse effects with calcium intakes as high as 2000 mg/d.

Hatton and McCarron [38] have reviewed the many possible mechanisms by which dietary calcium might affect blood pressure. One possible mechanism relates to a defect in calcium metabolism that leads to hypercalciuria. This defect would increase levels of parathyroid hormone, which in turn increases blood pressure; increased calcium intake would suppress the increased parathyroid hormone level and thus decrease blood pressure. Many other potential mechanisms have been discussed, including direct effects on vascular smooth muscle, calcium-regulating and calcium-sensitive hormones, interactions between calcium and sodium-potassium, and increased appetite for sodium when calcium intake is low. Further research in this area is needed.

Our pooled analysis suggested that calcium intake may have a larger blood pressure-decreasing effect in older persons and women. Further study may identify subgroups in which a higher calcium intake may have a clinically significant blood pressure-decreasing effect that warrants recommendations for increased calcium intake in these groups.

Our current overview does not support advising increased calcium intake for the primary prevention of hypertension. When the results of our metaanalysis of observational studies [8] are adjusted to the median dose among clinical trials (1 g of calcium), the blood pressure changes in observational studies and clinical trials are similar. Given these results, it does not seem appropriate to recommend changes in the current recommended daily allowance levels of 800 mg of calcium for men and nonpregnant women and 1200 mg for pregnant women. The best advice may be to consume calcium at these levels. (Slightly higher levels are recommended in the consensus statement [35].) Current recommended daily allowance levels, or those revised to improve maintenance of bone density, could be included in a comprehensive set of nutritional and lifestyle recommendations for cardiovascular health and in general nutritional guidelines.

Appendix

Let T_i denote the estimate of treatment effect in trial i, T_i = D_Ti − D_Ci, where D_Ti and D_Ci are the average changes in blood pressure between intervention and follow-up in the treatment and control groups, respectively. The variance of T_i is v_i = δ_i ² (1/n_Ti + 1/n_Ci) for parallel studies and v_i = δ_i ²/n_i for crossover studies. For parallel studies, δ_i ² is the variance of the changes in blood pressure between intervention and follow-up and n_Ti and n_Ci denote the number of patients in whom blood pressure was measured before and after intervention in the treatment and control groups, respectively. For crossover studies, δ_i ² is the variance of the blood pressure change over the treatment and control phases, and n_i is the sample size. Our weighted estimate of the pooled treatment effect is Equation 1 and the standard error of the pooled treatment effect is Equation 2

All trials reported T_i; however, only 7 reported an estimate of V_i for diastolic blood pressure, and only 10 reported an estimate for systolic blood pressure. For the studies that did not report v_i, we used a variance imputation in which δ_i ² was replaced with a weighted average of the reported δ_i ²S [39]. Selecting this method of variance imputation did not have a large effect on our results.