This systematic literature review with meta-analysis was conducted in accordance with the methodological guidelines of the Cochrane Handbook for Systematic Reviews of Interventions and reported in accordance with the PRISMA⁷ recommendations (Table S1 – Supplementary Material). The study protocol was registered on the PROSPERO platform on June 17, 2023 (CRD42023429824).

The research question was based on the PICOS framework - population (P), intervention (I), comparator (C), outcomes (O), and study design (S) - which were defined as follows:

Searches were conducted in the PubMed, EMBASE, and LILACS databases between July 1, 2022, and April 30, 2025. The search strategy included descriptors related to calcium-based medications and supplements and cardiovascular risks, which were combined using the Boolean operators OR and AND. Details of the search strategies used for each database are available in the supplementary material (Table S2).

In addition, a manual search of the bibliographic references of the selected studies was performed.

Primary observational and interventional studies that assessed cardiovascular risk in individuals using calcium-based medications or supplements were included.

The inclusion of both interventional and observational studies was justified by the interest in evaluating outcomes in controlled clinical settings through interventional studies, as well as in real-world settings, where patients were embedded in their social and healthcare contexts, through observational studies.

In addition, post hoc analyses of previously published pivotal studies were included, even if not specified in the protocol. This decision was made by mutual agreement among all members of the research team, as it was noted during the eligibility stage that several post hoc analyses met the research criteria and were therefore included in the study.

Articles published in non-Roman scripts (e.g., Chinese, Japanese) were excluded, as were studies in which the target population was not adults and/or in which the medication and/or supplement was administered via a route other than oral.

Articles identified through the searches were exported to EndNote^® software, where duplicates were removed. Subsequently, the records were uploaded to the Rayyan^® platform. Two authors independently selected the studies, following these steps: (1) screening (review of titles and abstracts) and (2) eligibility (full-text review). Any disagreements between the authors were resolved by consensus, and a third reviewer was involved when necessary.

The following study characteristics were extracted:

Regarding the study: study design and type, year of publication, country in which the study was conducted, and number of participants.

Regarding participant characteristics: data such as age, sex, ethnicity, comorbidities, type of therapy used and evaluated (medication or supplement use), as well as dosage, regimen, and duration of follow-up.

Regarding outcomes: associated cardiovascular risk, method used to assess cardiovascular risk, any other significant health conditions, and main study conclusions.

The data extraction process was performed independently by two authors, and discrepancies were resolved by a third reviewer. Microsoft Excel was used for data recording, extraction, and management.

The risk of bias assessment of the included studies was performed independently by two reviewers, with a consensus meeting held to compare and reconcile the assessments. Discrepancies were resolved by a third reviewer.

Primary observational studies were assessed using the Risk of Bias in Non-Randomized Studies of Interventions (ROBINS-I) tool, and interventional studies were assessed using the Risk of Bias Tool for Randomized Trials (RoB 2.0).

To assess the quality of the evidence, the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system was applied to each included outcome. The assessment was performed by two independent authors, and discrepancies were resolved through discussion with a third reviewer.

Statistical analysis (meta-analysis) was conducted using Review Manager^® (RevMan) software, based on data extracted from the studies.

Analyses were performed using a random-effects model, applying the Mantel–Haenszel method for dichotomous outcomes. The effect measure used was relative risk (RR) with a 95% confidence interval (CI) and a significance level of 0.05. Heterogeneity was analyzed using the I² value, and sensitivity was explored in analyses in which the I² value was above 50%, which is considered a high level of heterogeneity.⁷

These decisions were made and agreed upon by all researchers to standardize the analyses and reduce the risk of bias by proportionally weighting the contribution of each study to the meta-analysis, considering differences in populations and outcomes across studies.

Subgroup analyses were performed in populations specified in the protocol and included patients using calcium with or without vitamin D, as well as female patients.

The database search identified 3,276 records, of which 46 were duplicates, resulting in a total of 3,230 records. No eligible studies were identified through the manual search. None of the retrieved records was excluded due to non-Roman characters. After the screening process, 95 studies were selected for full-text review, of which 79 were excluded based on the eligibility criteria (Table S3 – Supplementary Material), and 2 were not available in the databases. At the end of the selection process, 14 studies were included for qualitative analysis as well as quantitative analysis (meta-analysis) (Figure 1).

Figure 1

Regarding interventional studies, seven articles were included, corresponding to three studies (Table S4 – Supplementary Material).

All seven were double-blind randomized clinical trials (RCTs), in which the interventions compared calcium in specific molecular forms with placebo.

All RCTs evaluated only female patients, with Bolland et al., 2008,⁸ Boursiquot et al., 2009,⁹ Cauley et al., 2013,¹⁰ and Hsia et al., 2007¹¹ restricting the population to postmenopausal women, that is, women in whom this condition had already been established and was present at the time of the study.

Regarding potential conflicts of interest, only the study by Hsia et al., 2007¹¹ reported receiving funding from a pharmaceutical company. The other studies reported no such funding.

The studies by Lewis et al., 2011,¹² Bolland et al., 2008,⁸ and Hsia et al., 2007¹¹ are pivotal studies, whereas the studies by Boursiquot et al., 2009,⁹ Cauley et al., 2013,¹⁰ Kato et al., 2023,¹³ and Prentice et al., 2013¹⁴ are post hoc studies related to the study by Hsia et al., 2007.¹¹

A 0.05 significance level was adopted for all included studies.

Regarding observational studies, seven articles were included (Table S5 – Supplementary Material).

The study by Godinho et al., 2021¹⁵ was the only Brazilian study and also the only cross-sectional study. The methodology was based on collecting data from women using calcium-based medications or supplements at the time of the study and whether they had reported the outcomes assessed. Because the data were collected at a specific point in the evaluation timeline, the study was included; however, it was not possible to perform a meta-analysis with the data obtained.

All six other studies were cohort studies, with Harvey et al., 2018,¹⁶ Li et al., 2012,¹⁷ and Pentti et al., 2009¹⁸ being prospective cohorts, and Kassis et al., 2022,¹⁹ Kim et al., 2022,²⁰ and Park et al., 2022²¹ being retrospective cohorts.

Three studies evaluated calcium in combination with vitamin D: Godinho et al., 2021,¹⁵ Kassis et al., 2022,¹⁹ and Kim et al., 2022.²⁰ The remaining studies did not mention the possibility of combining vitamin D in the supplements administered to participants.

The studies by Godinho et al., 2021¹⁵ and Pentti et al., 2009¹⁸ were the only ones to evaluate women exclusively, whereas the others evaluated both men and women.

Regarding potential conflicts of interest, only researchers in the study by Harvey et al., 2018¹⁶ declared having received funding from different pharmaceutical companies. The other studies reported not having received any.

A 0.05 significance level was adopted for all included studies.

The reported data were correlated with the outcome data analyzed in this review, and the final results are presented in the supplementary material (Tables S6 and S7).

For interventional studies, the main concern was related to domain 5, which refers to bias in the selection of reported outcomes. No studies were penalized for the other domains.

Regarding the risk of bias analysis for observational studies, penalties occurred in domains 1, 3, 5, and 6, referring, respectively, to bias due to confounding, bias in the classification of interventions, bias due to missing data, and bias in the measurement of outcomes.

Among the outcomes eligible for meta-analysis, four were included: mortality from cardiovascular events, occurrence of heart attacks, occurrence of strokes (ischemic stroke), and occurrence of CAD .

The meta-analyses were stratified according to study design. A subgroup analysis was performed including patients using calcium combined with vitamin D. The analysis of pre-established health conditions, as proposed in the protocol, was not performed due to a lack of available data in the included studies.

Furthermore, in meta-analyses involving interventional studies, priority was given to including only the pivotal study rather than post hoc analyses, whenever possible, in accordance with studies derived from the same pivotal study.

The results obtained are presented in Figure 2 (interventional studies) and Figure 3 (observational studies).

Figure 2

Figure 3

A meta-analysis was also performed considering only the subgroup of observational studies in which the intervention evaluated was the use of calcium-based supplements combined with vitamin D, as shown in Figure 4. Two studies were evaluated, which demonstrated no difference between the group of patients using calcium-based and vitamin D supplements and those not using supplements (RR 1.30; 95% CI [0.46; 3.67]), although with high heterogeneity (I² = 93%; p = 0.0002). Sensitivity analyses were not possible because only two studies were analyzed.

Figure 4

It was possible to perform three sensitivity analyses in meta-analyses presenting high heterogeneity (I² > 50%), namely: (i) the outcome of cardiovascular event mortality in interventional studies, (ii) the outcome of cardiovascular event mortality in observational studies, and (iii) the outcome of myocardial infarction occurrence in observational studies.

Among these, for outcome (ii), the removal of the study by Kim et al.,²⁰ 2022 was identified as responsible for the observed heterogeneity value. The study was reviewed to investigate potential clinical, methodological, and/or statistical variability, and one possible difference identified was that the study population was of Asian origin, whereas the remaining studies primarily included Caucasian populations. No other relevant differences were identified upon review of the study.

For outcome (iii), the study by Li et al., 2012,¹⁷ was identified as responsible for the observed heterogeneity. The study was reviewed to assess potential clinical, methodological, and/or statistical variability, and one possible difference was that the study population was of Caucasian origin, whereas the other two studies primarily included Asian populations. No additional differences were identified during the review of the study.

Furthermore, additional subgroup analyses or sensitivity analyses were not feasible due to the insufficient number of records (more than one) required to support the development of statistical analyses.

The results of the quality of evidence assessment generated using the GRADE tool²² were grouped and are presented in Table S8 for intervention studies and Table S9 for observational studies, available in the supplementary material.

Penalties occurred in three of the tool’s domains: (i) related to risk of bias: a severe penalty was applied due to studies included in the outcomes presenting bias in at least one assessment domain; (ii) related to inconsistency: the outcomes were penalized for presenting heterogeneity, considering high I² values and statistically significant p values, as well as studies that did not share a CI with the diamond; and (iii) related to imprecision: a penalty was applied for CIs greater than or less than 50% of the baseline and for non-symmetry of the final results of the diamonds obtained.

The central figure represents a summary of the main findings of this article.