点击显示 收起
1 From the Departments of Nutrition (HAB-F, EO, and WCW), Epidemiology (RL and DS), and Biostatistics (RL, DS, and JEO), Harvard School of Public Health, Boston, MA; the Department of Rheumatology and the Institute for Physical Medicine and Rehabilitation, University Hospital Zurich, Zurich, Switzerland (HAB-F); the Jean Mayer USDA Human Nutrition Research Center on Aging, Tufts University, Boston, MA (BD-H); the Departments of Medicine and of Community and Family Medicine, Dartmouth Medical School, Hanover, NH (JAB); Clinique Bois Cerf/Hirslanden, Lausanne, Switzerland (PB); the EA Martin Program in Human Nutrition, South Dakota State University, Brookings, SD (BS); the Division of Clinical Decision Making, Tufts-New England Medical Center, Boston, MA (JBW); the Department of Geriatrics, University of Basel, Basel, Switzerland (HBS); and the Institute for Aging Research, Hebrew SeniorLife, Harvard Medical School, Boston, MA (DPK)
See corresponding editorial on page 1579.
ABSTRACT
Background: The role of total calcium intake in the prevention of hip fracture risk has not been well established.
Objective: The objective of the study was to assess the relation of calcium intake to the risk of hip fracture on the basis of meta-analyses of cohort studies and clinical trials.
Results: In women (7 prospective cohort studies, 170 991 women, 2954 hip fractures), there was no association between total calcium intake and hip fracture risk [pooled risk ratio (RR) per 300 mg total Ca/d = 1.01; 95% CI: 0.97, 1.05]. In men (5 prospective cohort studies, 68 606 men, 214 hip fractures), the pooled RR per 300 mg total Ca/d was 0.92 (95% CI: 0.82, 1.03). On the basis of 5 clinical trials (n = 5666 women, primarily postmenopausal, plus 1074 men) with 814 nonvertebral fractures, the pooled RR for nonvertebral fractures between calcium supplementation (800–1600 mg/d) and placebo was 0.92 (95% CI: 0.81, 1.05). On the basis of 4 clinical trials with separate results for hip fracture (6504 subjects with 139 hip fractures), the pooled RR between calcium and placebo was 1.64 (95% CI:1.02, 2.64). Sensitivity analyses including 2 additional small trials with <100 participants or per-protocol results did not substantially alter results.
Conclusions: Pooled results from prospective cohort studies suggest that calcium intake is not significantly associated with hip fracture risk in women or men. Pooled results from randomized controlled trials show no reduction in hip fracture risk with calcium supplementation, and an increased risk is possible. For any nonvertebral fractures, there was a neutral effect in the randomized trials.
Key Words: Meta-analysis • hip fracture • nonvertebral fracture • calcium intake • calcium supplementation • cohort studies • randomized controlled trials
INTRODUCTION
Calcium supplementation or the consumption of calcium-rich foods such as milk is commonly recommended for the prevention of osteoporosis and fractures. These recommendations are primarily based on evidence from randomized controlled trials (RCTs) with bone density as the outcome. However, in a 2004 meta-analysis of RCTs, supplementation with 500–2000 mg Ca/d in postmenopausal women provided only a modest benefit for bone density: 2.05% difference in total-body bone density, 1.66% difference in lumbar spine bone density, and 1.64% difference in hip bone density (1, 2). The implications of such differences with respect to fracture risk prevention are unclear. In the same meta-analysis, limited evidence from RCTs (222 subjects in 2 trials) suggested only a modest and nonsignificant benefit of calcium supplementation for the risk of nonvertebral fractures [pooled risk ratio (RR) = 0.86; 95% CI: 0.43, 1.72]. Furthermore, an earlier meta-analysis published in 1997 that summarized observational studies in postmenopausal women found no clear benefit of a 300-mg increment in daily calcium intake for hip fracture risk [pooled RR among 28 511 women from 5 cohorts was 0.96 (95% CI: 0.91, 1.02)] (3). Consequently, considerable uncertainty exists regarding optimal intakes of calcium, and this uncertainty is reflected in markedly different recommended daily intakes among countries. For example, for adults >50 y old, the recommended daily intake is 700 mg Ca/d in the United Kingdom and 1200 mg Ca/d in the United States (4).
Several RCTs tested the combined effect of calcium plus vitamin D, and this evidence was summarized in 2 recent meta-analyses that suggested, irrespective of trial quality, a small but significant reduction in hip fracture risk [pooled RR = 0.81; 95% CI: 0.68, 0.96 (5); pooled RR = 0.82; 95% CI: 0.71, 0.94 (6)]. However, the benefit of calcium supplementation alone was not addressed in these analyses. Because increased calcium intake alone is still commonly recommended as a fracture-prevention strategy (6, 7), the assessment of calcium intake and its effect on hip fracture risk reduction is of clinical importance.
Thus, we conducted a systematic review and meta-analysis of prospective cohort studies to address these relations with respect to hip fracture prevention. Hip fractures are the most severe and the most frequent fractures in older persons (8, 9). We focused on prospective cohort studies because they are less susceptible to selection and recall biases than are case-control studies. Because several RCTs of calcium supplementation without vitamin D have been conducted since the mid-1990s, and all have had samples sizes that were somewhat small for the assessment of hip fractures (10–14), we also summarized findings regarding the effect of calcium supplementation on all nonvertebral fractures from randomized trials.
METHODS
Search strategy and data extraction
For both prospective cohort studies and RCTs, we conducted a systematic search for relevant English and non-English publications by using MEDLINE (Ovid and PubMed) for the period from January 1960 to December 2006 and by using EMBASE for January 1991 to December 2006. We also contacted experts in the field and searched reference lists and abstracts presented at the meetings of the American Society for Bone and Mineral Research from 1995 through 2006.
For prospective cohort studies, we used numerous medical subject headings (MeSH terms)—"cohort studies" or "prospective studies," or "retrospective studies," and "fracture," or "hip fracture," and "calcium," or "calcium analogs or derivates," or "calcium carbonate," or "calcium citrate," or "calcium gluconate," or "calcium phosphate," or "milk," or "dairy products." To update the most recent meta-analysis of RCTs on calcium and fracture risk, we also searched for RCTs and fracture risk by using the MeSH terms above plus "randomized-controlled trial" or "controlled-clinical trial" or "random allocation" or "double-blind method" or "uncontrolled trial." In addition, we searched for fracture data in trials that had bone density as the primary outcome. We received unpublished data from one large published trial for calcium supplementation, the Randomised Evaluation of Calcium Or vitamin D (RECORD) trial (12)].
Eligibility and exclusion criteria were specified in advance. Data extraction was conducted independently by 2 investigators (HAB-F and EO).
Eligible studies
For prospective cohort studies, we included only studies in which calcium intake had been assessed before the fracture events. Our primary outcome was the first incident hip fracture in middle-aged or older men and women. For trials that addressed fracture incidence, we included only double-blind RCTs that studied any dose of calcium supplementation compared with placebo. Because of limited data on hip fractures, in separate analyses, we also included all nonvertebral fractures. We included only double-blind RCTs that studied calcium supplementation with a minimum follow-up of 1 y and that required >100 study participants. Trials with <100 participants were added in a sensitivity analysis.
Ineligible studies
We excluded uncontrolled trials, cross-sectional and case-control studies, and animal investigations. Of prospective cohort studies, we excluded studies that did not provide separate data for men and women (15) or for hip fracture (16). Of RCTs, we excluded studies in which calcium was combined with other agents, such as vitamin D (17–19), because the effects of the two agents could not be separated. The combination of calcium plus vitamin D has been addressed in 2 recent meta-analyses of RCTs (5, 6).
Studies identified
Prospective cohort studies
A total of 8 separate studies were identified—7 that included women (3, 20–25) and 5 that included men (20–23, 26) (Table 1 and Figure 1 A). Total calcium intake included dietary and supplement sources in 4 studies (3, 21, 25, 26) and only dietary calcium intake in 4 studies (20, 22–24). This omission may not have been important in the older studies (20, 23), because calcium intake from supplements became widespread only in the late 1980s (21).
View this table:
TABLE 1. Prospective cohort studies that assessed total calcium intake or milk intake and hip fracture risk1
FIGURE 1.. A: A Quality of Reporting of Meta-analyses (QUOROM) flow diagram for the prospective cohort studies. B: A QUOROM flow diagram for the randomized controlled trials (RCTs).
Randomized controlled trials
We identified 5 RCTs that met our criteria for the primary analysis (10, 12–14, 27) and 2 additional smaller trials that were included in the sensitivity analysis (11, 28). RCTs primarily included postmenopausal women (Table 2). Five of these 7 RCTs provided separate data on hip fractures (10–14). Our analyses followed the intention-to-treat principle. Only 3 trials provided a per-protocol analysis(10, 12, 13). Despite limited data for men, results by sex were pooled for the primary intention-to-treat analysis examining both hip and any nonvertebral fractures.
View this table:
TABLE 2. Randomized controlled trials of calcium supplementation and fracture (fx) risk1
Statistical methods
Sex-specific cohort study analyses were conducted because men and women differ in fracture risk (29) and calcium intake (30). The primary outcome of the pooled analysis was the RR of hip fracture for a 300-mg increment in daily calcium intake, the amount of calcium in an 8-ounce (237-mL) glass of milk or one slice (50 g) of hard cheese.
For the highest and the lowest open-ended calcium intake categories, we chose a previously defined value for a corresponding median—30% lower than the lowest cutoff and 30% higher than the upper cutoff. RRs adjusted for multiple covariates were used whenever available. The study of Holbrook et al (20) provided, for each sex, the RR, the overall sample size, and the overall number of hip fractures. From these numbers, we calculated the number of hip fracture cases in each exposure group and re-ran the analyses of the 2 x 2 tables to retrieve the corresponding 95% CIs for the RRs.
To compare studies on the same scale in the pooled analysis, we calculated the RR for a 300-mg increment in total daily calcium intake, which assumes a log-linear association of intake with risk. Because the relative risks within each cohort study depend on a common reference group, they are correlated. Thus, we used a method developed by Greenland et al (31), which yields an efficient point estimator and a consistent variance estimate under these circumstances, to calculate for each study the RR of hip fracture per 300-mg increase in total daily calcium intake. Results from all studies were then pooled by using random-effects models (32).
RCT outcomes were pooled on an intention-to-treat basis with random-effects models, which control for both within-trial and between trial variance. Heterogeneity among RCTs and among cohort studies was evaluated by using the Q statistic, which is considered significant for P < 0.10 (33, 34). To address the effect of poor adherence in some larger trials, a sensitivity analysis was performed on the basis of the per-protocol results of those trials.
To assess potential publication bias, we used the Begg and Egger tests and Begg's funnel plot (35, 36); no evidence of bias was seen in the prospective cohort studies or the RCTs. Statistical analysis was performed by using STATA software (version 7.0; Stata Corp, College Station, TX).
RESULTS
Prospective cohort studies
Characteristics of the 8 prospective cohort studies that met our inclusion criteria are shown in Table 1. Of these 8 studies, 7 included 170 991 women who sustained 2954 hip fractures, and 5 included 68 606 men who sustained 214 hip fractures. The median age at baseline ranged from 41 to 72 y. Mean follow-up varied between 3 and 18 y. Six studies were from the United States (3, 20–22, 25, 26), one was from Norway (23), and one was from Sweden (24).
Primary analysis
The Forest plots for the RR of hip fracture for a 300-mg increase in daily calcium intake are shown in Figure 2. In women, there was no association between total calcium intake and hip fracture risk (pooled RR for additional 300 mg Ca/d intake = 1.01; 95% CI: 0.97, 1.05). In men, the pooled RR was 0.92 (95% CI: 0.82, 1.03).
FIGURE 2.. Relative risk (RR) of hip fracture for a 300-mg increase in total calcium intake per day from the prospective cohort studies. The size of the squares is proportional to the inverse of the variance. Error bars represent the 95% CIs. The confidence limits for the pooled RR are indicated by the diamond-shaped figure. There was no significant heterogeneity among studies in men or women (P = 0.34 and 0.55 for women and men, respectively; Q test). A: In the women, the pooled RR per 300 mg total Ca/d was 1.01 (95% CI: 0.97, 1.05). B: In the men, the pooled RR per 300 mg total Ca/d was 0.92 (95% CI: 0.82, 1.03).
Subgroup analyses for hip fracture risk and total calcium intake
In the 4 studies of women that reported calcium intake from dietary sources alone, ie, ignoring supplemental calcium (20, 22–24), we found no association between dietary calcium intake and risk of hip fracture (pooled RR per additional 300 mg Ca/d intake from dietary calcium = 1.01; 95% CI: 0.96, 1.06). In the 2 studies that measured dietary calcium intakes by using 24-h recall (20, 22), the association between calcium intake and hip fracture risk (pooled RR per additional 300 mg Ca/d intake = 0.90; 95% CI: 0.76, 1.07) was somewhat stronger than that in the 5 studies that used food-frequency questionnaires (pooled RR per additional 300 mg dietary Ca/d calcium = 1.01; 95% CI: 0.98, 1.05).
Because the efficacy of calcium may be enhanced by additional vitamin D, as found in RCTs involving both institutionalized (17) and ambulatory (19) women, we sorted studies by latitude (south to north), taking the mean of the state capitals of the multistate US studies (Figure 2A). This analysis showed no stronger protective effect of calcium intake on hip fracture risk in women living in southern latitudes, who possibly had higher vitamin D status due to increased sunshine exposure, than in women living in northern latitudes. The limited data for men did not allow useful subgroup analyses.
To examine the relation between calcium intake and hip fracture risk in more detail, we pooled RRs for categories of total calcium intake and hip fracture risk among women from each cohort study (Figure 3), using the lowest category as a reference and corresponding RRs for higher intake categories. Figure 3 confirms the findings suggested by the analysis in Figure 2A—ie, there is no apparent association between calcium intake and hip fracture risk over a wide range of calcium intakes in women. The limited data for men did not allow useful categorical dose-response analyses.
FIGURE 3.. Pooled analysis for categories of calcium intake and hip fracture risk in the women from the prospective cohort studies. The reference intake categories in the various studies ranged from 280 to 554 mg total Ca/d.
Randomized controlled trials
Characteristics of the 5 RCTs that met our criteria for the primary analysis(10,12–14, 27) and 2 smaller trials that were included in the sensitivity analysis (11, 28) are shown in Table 2. The mean age ranged from 58 to 77 y. The mean duration of follow-up varied between 1.5 and 10.8 y. One study was from Switzerland (28), 1 study was from the United Kingdom (12), 2 studies were from New Zealand (10, 11), 2 were from the United States (14, 27), and 1 was from Australia (13).
The primary analysis for nonvertebral fracture risk included 5 RCTs (Table 3; Figure 4) in which calcium supplementation between 800 and 1600 mg/d was compared with placebo. Among 6740 subjects (5666 women, primarily postmenopausal, plus 1074 men) who had a total of 814 nonvertebral fractures, the pooled RR was 0.92 (95% CI: 0.81, 1.05). This result did not change substantially in the sensitivity analyses: including the 2 additional small trials, the pooled RR was 0.91 (95% CI: 0.80, 1.03), and including only adherent subjects, the pooled RR was 0.83 (95% CI: 0.64, 1.09).
View this table:
TABLE 3. Evidence table for randomized controlled trials comparing calcium supplementation with placebo for risk of all nonvertebral fractures (fxs)1
FIGURE 4.. Forest plot comparing intention-to-treat data from 5 randomized controlled trials for the risk of nonvertebral fractures between calcium-treated and placebo groups. The squares represent the relative risk (RR) of fracture between subjects who took calcium in any dose and those who took placebo. In a total of 6740 subjects, the pooled RR was 0.92 (95% CI: 0.81, 1.05). There was no heterogeneity between studies (P = 0.77, Q test).
Intention-to-treat results comparing calcium with placebo in women alone (pooled RR = 0.92; 95% CI: 0.81, 1.06) were similar to the results including men. The pooled RR for men from 2 trials was 0.94 (95% CI: 0.64, 1.37) (12, 14).
The primary analysis for hip fracture risk included 4 RCTs (Table 4; Figure 5) in which calcium supplementation between 800 and 1200 mg/d was compared with placebo. Among 6504 subjects (5430 women, primarily postmenopausal, plus 1074 men) who had a total of 139 hip fractures, the pooled RR was 1.64 (95% CI: 1.02, 2.64). This result did not change significantly in sensitivity analyses: including the 2 additional small trials, the pooled RR was 1.57 (95% CI: 0.96, 2.55), and including only adherent subjects, the pooled RR was 1.42 (95% CI: 0.81, 2.49). Intention-to-treat results comparing calcium with placebo in women alone (pooled RR = 1.66; 95% CI: 0.97, 2.86) were similar to the results including men. The pooled RR for men from 2 trials was 1.55 (95% CI: 0.62, 3.96) (12, 14).
View this table:
TABLE 4. Evidence table for randomized controlled trials comparing calcium supplementation with placebo for risk of hip fractures (fxs)1
FIGURE 5.. Forest plot comparing intention-to-treat data from 4 randomized controlled trials for the risk of hip fracture between calcium-treated and placebo groups. The squares represent the relative risk (RR) of fracture between subjects who took calcium in any dose and those who took placebo. In a total of 6504 subjects, the pooled RR was 1.64 (95% CI: 1.02, 2.64). There was no heterogeneity between studies (P = 0.24, Q test).
DISCUSSION
In our meta-analysis of prospective cohort studies, calcium intake was not significantly associated with hip fracture risk in men or women. Similarly, our meta-analysis of RCTs, which included data largely from postmenopausal women, yielded a neutral effect of calcium supplementation as compared with placebo for any nonvertebral fracture and suggested a significantly (64%) greater risk of hip fractures with calcium supplementation.
There are several possible explanations for the lack of overall association between total calcium intake and hip fracture risk in prospective cohort studies. Calcium intake is imperfectly measured in observational studies, and this measurement error—if nondifferential—would lead to an underestimation of a true calcium effect. However, in several of the studies, the validity of the calcium intake estimates was assessed by comparison with more detailed methods (37, 38), and the correlations between food-frequency questionnaires and 1-wk diet records were high, 0.75 (37, 39). This modest degree of measurement error would tend to cause a conservative bias, but, with the large number of cases in women than in men, an important association should not have been missed. Moreover, in some of the same studies that reported fracture data, dietary calcium intake has been inversely associated with the risk of kidney stones (38, 40, 41) and colon cancer (42), which showed that the calcium intake measures used were at least accurate enough to detect those relations.
Another possible explanation for the lack of association is that patients with recognized osteoporosis are generally advised to increase calcium intake, which could mask an inverse association between calcium intake and fracture risk (29). However, in the Nurses' Health Study, the exclusion of women with a history of diagnosed osteoporosis did not appreciably affect its negative findings (25).
Alternatively, calcium alone may not prevent hip fractures in women. In fact, our meta-analysis of RCTs suggests an increased risk with calcium supplementation among men and women. It is possible that, among the frail subjects at risk of hip fracture, other deficiencies, such as vitamin D deficiency and phosphate deficiency due to low protein intake, should be corrected along with ensuring adequate calcium intake (17). Calcium carbonate or calcium citrate supplements can reduce phosphate absorption (43), which may be detrimental, because a balanced ratio of calcium to phosphate is needed for bone mineralization (44). Phosphate deficiency [defined as an intake <70% of adult recommended dietary allowance (700 mg/d)] is found in 10% of US women >60 y old and in 15% of US women >80 y old (45). Each increase in calcium intake by 500 mg/d decreases phosphorus absorption by 166 mg (43), so a calcium supplement of 1000 mg may shift an elderly person with a relatively low phosphorus intake into phosphate deficiency (43, 46). This change could augment bone resorption (43, 47, 48) and thus increase fracture risk. Conversely, in the trial by Chapuy et al (17), the beneficial effect of vitamin D plus calcium on hip fracture risk in frail elderly women may have been enhanced by the use of tricalcium phosphate, which may have avoided a calcium-related phosphate deficiency. Furthermore, vitamin D stimulates phosphate absorption (49), which may enhance phosphate uptake from nutritional sources in calcium supplement users. Such a benefit is supported by a recent meta-analyses showing that hip fracture risk is significantly reduced in trials that combined any calcium supplement with vitamin D (5, 6). Similarly, the main dietary sources of calcium also contain phosphorus, which could explain the lack of a positive association between calcium intake and fracture risk in the cohort studies.
An alternative explanation for the pooled RCT data on hip fractures is poor statistical power. The 4 studies that were pooled for the hip fracture outcome were limited by a relatively small number of cases (139 fractures), and one trial had poor adherence and a focus on secondary fracture prevention (12). Thus the apparent elevation in risk may be due to chance. However, if we considered only adherent subjects from the 3 largest trials, the elevated risk was maintained. Furthermore, each of the 4 trials included in the primary analysis indicated an elevated risk of hip fracture with RRs between 1.21 and 3.43, although only 1 of these trials reached significance (10). Thus the consistency of the evidence argues against a chance finding. Moreover, as discussed above, commonly used calcium supplements with a carbonate or citrate component have been shown to shift older persons into phosphate deficiency (43, 46) and to induce bone loss (43, 47, 48). To build calcium into bone, a calcium-phosphate product is needed, which may be disturbed by the described calcium supplement–induced phosphate malabsorption. This may be especially critical in frail older persons ndividuals at risk of both hip fracture and phosphate deficiency because of low protein intake. Thus, our findings are mechanistically plausible.
It is most important, however, that, with a CI excluding 1 for the primary analysis, an important reduction in hip fracture risk with calcium supplementation seems unlikely. Furthermore, despite the limited data on men, a differential effect of calcium on hip fracture risk by sex appears unlikely: the pooled RRs were 1.66 among women and 1.56 among men. Still, more data on fracture risk in men are needed.
The strengths of our meta-analysis of prospective cohort studies are the large number of cases, the long duration of follow-up, and the inclusion of both men and women. Prospective cohort studies have less potential for bias than do other observationally designed studies, because the data on calcium intake are assessed before occurrence of fractures. In the presence of limited data from RCTs, as confirmed in our meta-analysis, summarization of these studies is likely to be the most informative approach.
However, our analysis has limitations. Prospective cohort studies may still be susceptible to bias, including loss to follow-up and residual confounding. Another limitation of our study is that the calcium intake from supplements was not assessed in all of the studies. Nonetheless, we did not detect heterogeneity between studies that assessed calcium intake from both dietary and supplement sources and those that studied only dietary intake. Our study was also limited by the lack of information on baseline 25-hydroxyvitamin D concentrations, phosphate intake, and physical activity. These factors could potentially modify the associations between calcium intake and fracture risk (17, 19, 50). In the RCTs included in our meta-analysis, participants' mean baseline calcium intakes exceeded the estimated mean intake of persons aged 50 y in the general US population—ie, 763 mg/d in men and 558 mg/d in women (22). Thus, we cannot exclude the possibility that persons with very low baseline calcium intakes may benefit more from calcium supplementation than those with higher calcium intakes; another reason for the negative findings could be that the subjects in the RCTs pooled for this meta-analysis already had "enough" calcium.
In summary, our results do not support an overall beneficial effect of greater calcium intake on hip fracture risk. Among women, the cohort data suggest a neutral effect of calcium intake on hip fractures, but data from RCT's of calcium supplementation suggest an adverse effect, even among adherent women. In addition, RCT data for any nonvertebral fractures indicate a neutral effect of calcium with respect to fracture reduction. Thus, future studies of the prevention of hip fracture or any nonvertebral fracture in women should not consider calcium supplementation alone but, rather, should focus on the optimal combination of calcium plus vitamin D and possibly on the correction of phosphate deficiency by using calcium-phosphate supplements. RCT and prospective cohort study findings for men did not support a beneficial effect of calcium intake on hip fracture risk in men, but further studies in men are needed as data are limited.
ACKNOWLEDGMENTS
The authors thank the investigators of the RECORD trial for providing unpublished data for this meta-analysis.
The authors' responsibilities were as follows—HAB-F had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis; HAB-F, WCW, JAB, and BD-H: study concept and design; HAB-F and EO: acquisition of data; all authors: analysis and interpretation of data and critical review of the manuscript; HAB-F, WCW, JAB, BD-H, JEO: writing the manuscript draft; HAB-F, WCW, JEO, DS, and RL: statistical analysis; HAB-F: obtained funding; and WCW: administrative, technical, or material support. JBW receives funding from federal agencies, Schering Plough, and Centocor for work unrelated to studies of calcium or falls and fractures; JAB receives study agents (medications) from Wyeth, and, with Dartmouth College, holds a use patent for calcium as a cancer chemopreventive agent. No other authors had any personal or financial conflict of interest.
REFERENCES
Related articles in AJCN: