Calcium Intake and Risk of Cardiovascular Disease
Calcium Intake and Risk of Cardiovascular Disease
We searched the published literature for studies on calcium and CVD events between 1966 to 1 September 2010, using MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials. The search terms for calciumincluded calcium intake, calcium supplement, calcium carbonate, and calcium citrate. The search terms for CVD included cardiovascular disease, ischemic heart disease, coronary artery disease (CAD), cardiovascular mortality, myocardial infarction (MI), and stroke. The literature search was restricted to English-language articles, human studies, and adult subjects aged ≥19 years. The same search strategy was applied to each database. We also manually searched reference lists of the identified articles for additional studies.
Two investigators (Drs Wang and Sesso) independently reviewed all identified articles. We excluded ecological, crosssectional, and retrospective case-control studies due to concerns for temporality and reverse causation. Based upon reviewof the abstract, we also excluded case reports, studies of children and adolescents, and studies that did not assess levels of dietary calcium, calcium supplement use, or serum or plasma calcium levels. Articles that passed abstract screening were retrieved for a full-text review. We then further excluded review articles, editorials, the studies that did not compare risk of CVD between different levels of dietary, supplemental, or circulating calcium and the studies that did not ascertain CVD events including CVD death, non-fatal CAD or MI, and nonfatal stroke. One investigator (Dr Wang) extracted key information from selected studies to summary tables. Data extracted for observational studies included country, study design, sample size, participant characteristics (including age, gender, and comorbid conditions), dietary, supplemental, or blood calcium examined, CVD endpoints, follow-up period, andmain study findings. Data extracted for randomized trials included country, sample size, participant characteristics, dosage and duration of calcium supplement, follow-up period, primary endpoint(s), and main findings for CVD outcomes. For separate publications from the same study population, we considered each publication as an individual study. We did not contact authors for missing data.
For each individual study identified from the literature search, we extracted the reported relative risks (RRs) and 95% confidence intervals (CIs) of CVD events across categories of dietary calcium or calcium supplement use. When theRRs from individual studies could be combined, we conducted metaanalyses. The RRs used in the meta-analysis were those adjusted for covariates as comprehensively as possible. For the studies that did not report RRs but reported CVD event rates, we calculated the unadjusted RRs and 95% CIs.
When the RRs from individual studies could be combined, we conducted meta-analyses. The pooledRRand 95%CIs were calculated by the random-effects method of DerSimonian and Laird using STATA statistical software (version 7.0, STATA Corp., College Station, TX, USA).
2. Methods
2.1 Data Sources and Literature Search
We searched the published literature for studies on calcium and CVD events between 1966 to 1 September 2010, using MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials. The search terms for calciumincluded calcium intake, calcium supplement, calcium carbonate, and calcium citrate. The search terms for CVD included cardiovascular disease, ischemic heart disease, coronary artery disease (CAD), cardiovascular mortality, myocardial infarction (MI), and stroke. The literature search was restricted to English-language articles, human studies, and adult subjects aged ≥19 years. The same search strategy was applied to each database. We also manually searched reference lists of the identified articles for additional studies.
2.2 Study Selection and Data Extraction
Two investigators (Drs Wang and Sesso) independently reviewed all identified articles. We excluded ecological, crosssectional, and retrospective case-control studies due to concerns for temporality and reverse causation. Based upon reviewof the abstract, we also excluded case reports, studies of children and adolescents, and studies that did not assess levels of dietary calcium, calcium supplement use, or serum or plasma calcium levels. Articles that passed abstract screening were retrieved for a full-text review. We then further excluded review articles, editorials, the studies that did not compare risk of CVD between different levels of dietary, supplemental, or circulating calcium and the studies that did not ascertain CVD events including CVD death, non-fatal CAD or MI, and nonfatal stroke. One investigator (Dr Wang) extracted key information from selected studies to summary tables. Data extracted for observational studies included country, study design, sample size, participant characteristics (including age, gender, and comorbid conditions), dietary, supplemental, or blood calcium examined, CVD endpoints, follow-up period, andmain study findings. Data extracted for randomized trials included country, sample size, participant characteristics, dosage and duration of calcium supplement, follow-up period, primary endpoint(s), and main findings for CVD outcomes. For separate publications from the same study population, we considered each publication as an individual study. We did not contact authors for missing data.
2.3 Data Synthesis and Analysis
For each individual study identified from the literature search, we extracted the reported relative risks (RRs) and 95% confidence intervals (CIs) of CVD events across categories of dietary calcium or calcium supplement use. When theRRs from individual studies could be combined, we conducted metaanalyses. The RRs used in the meta-analysis were those adjusted for covariates as comprehensively as possible. For the studies that did not report RRs but reported CVD event rates, we calculated the unadjusted RRs and 95% CIs.
When the RRs from individual studies could be combined, we conducted meta-analyses. The pooledRRand 95%CIs were calculated by the random-effects method of DerSimonian and Laird using STATA statistical software (version 7.0, STATA Corp., College Station, TX, USA).
Source...