Evaluation of the Incremental Value of a Coronary Artery Calcium Score Beyond Traditional Cardiovascular Risk Assessment: A Systematic Review and Meta-analysis

Key Points

Question  Is there an incremental gain from the addition of a coronary artery calcium score (CACS) to a standard cardiovascular disease (CVD) risk calculator?

Findings  In this systematic review and meta-analysis, the pooled gain in C statistics from adding CACS was 0.036. Most participants reclassified as being at intermediate or high risk by CACS did not have a CVD event during follow-up (range, 5.1 to 10.0 years).

Meaning  Although CACS appears to add some further discrimination to standard CVD risk calculators, no evidence suggests that this provides clinical benefit.

Importance  Coronary artery calcium scores (CACS) are used to help assess patients’ cardiovascular status and risk. However, their best use in risk assessment beyond traditional cardiovascular factors in primary prevention is uncertain.

Objective  To find, assess, and synthesize all cohort studies that assessed the incremental gain from the addition of a CACS to a standard cardiovascular disease (CVD) risk calculator (or CVD risk factors for a standard calculator), that is, comparing CVD risk score plus CACS with CVD risk score alone.

Evidence Review  Eligible studies needed to be cohort studies in primary prevention populations that used 1 of the CVD risk calculators recommended by national guidelines (Framingham Risk Score, QRISK, pooled cohort equation, NZ PREDICT, NORRISK, or SCORE) and assessed and reported incremental discrimination with CACS for estimating the risk of a future cardiovascular event.

Findings  From 2772 records screened, 6 eligible cohort studies were identified (with 1043 CVD events in 17 961 unique participants) from the US (n = 3), the Netherlands (n = 1), Germany (n = 1), and South Korea (n = 1). Studies varied in size from 470 to 5185 participants (range of mean [SD] ages, 50 [10] to 75.1 [7.3] years; 38.4%-59.4% were women). The C statistic for the CVD risk models without CACS ranged from 0.693 (95% CI, 0.661-0.726) to 0.80. The pooled gain in C statistic from adding CACS was 0.036 (95% CI, 0.020-0.052). Among participants classified as being at low risk by the risk score and reclassified as at intermediate or high risk by CACS, 85.5% (65 of 76) to 96.4% (349 of 362) did not have a CVD event during follow-up (range, 5.1-10.0 years). Among participants classified as being at high risk by the risk score and reclassified as being at low risk by CACS, 91.4% (202 of 221) to 99.2% (502 of 506) did not have a CVD event during follow-up

Conclusions and Relevance  This systematic review and meta-analysis found that the CACS appears to add some further discrimination to the traditional CVD risk assessment equations used in these studies, which appears to be relatively consistent across studies. However, the modest gain may often be outweighed by costs, rates of incidental findings, and radiation risks. Although the CACS may have a role for refining risk assessment in selected patients, which patients would benefit remains unclear. At present, no evidence suggests that adding CACS to traditional risk scores provides clinical benefit.

In 2018, the US Preventive Services Task Force (USPSTF) concluded that available evidence was insufficient to assess the balance of benefits and harms of adding a coronary artery calcium score (CACS) to traditional risk assessment for cardiovascular disease (CVD) in asymptomatic adults to prevent CVD events.¹ The USPSTF noted that adding a CACS may improve calibration, discrimination, and reclassification, but the clinical meaning of this improvement was unknown. The American Heart Association recommends a CACS for CVD risk screening if it provides information to modify treatment decisions,² and the European Society of Cardiology recommends that a CACS may be considered in individuals at low or moderate risk.³ Three randomized clinical trials⁴ have evaluated the potential clinical impact of a CACS in asymptomatic adults; one trial with 450 participants⁵ found no significant improvements in modifiable cardiovascular risk factors at 1 year; another trial of 1934 participants⁶ found decreases in risk factors at 4 years; and a third trial of 56 participants (postmenopausal women)⁷ found risk factors declined less in the CACS-screened group at 1 year. No trials have found a difference in actual CVD events based on results of coronary artery calcium (CAC) scanning.¹ Despite these uncertainties about clinical benefits, CAC scans have become widely used in clinical practice.⁸

A few potential harms have been noted. First is the exposure to radiation. Based on American College of Radiology registry data, the estimated exposure is 1.7 mSV for a CAC scan, which is 17 times higher than that from a 2-view chest radiograph and 4.5 times higher than that from a mammogram⁹ (range, 0.4-2.1 mSV).¹ Second is the potential for adverse psychological effects from the patient being told their CAC scan results are positive.¹⁰ Third is the potential for CAC scan results to cascade a subsequent series of further cardiac tests and interventions that are not necessary in individuals who are otherwise stable clinically¹¹ or other noncardiac tests and interventions because of incidental findings such as lung nodules.¹² There are also costs to patients and the health care system from the CAC scan and cascaded tests and treatments.

The USPSTF reviewed evidence until May 22, 2017, but more studies have been published on the topic since then. We therefore undertook a new systematic review and meta-analysis to evaluate the incremental value of CAC scans above and beyond traditional cardiovascular risk assessment.

This systematic review and meta-analysis was not registered. The review protocol, template data collection forms, data extracted from included studies, data used for all analyses, and analytic code are available on request from the corresponding author. The secondary outcomes were not prespecified in the protocol. This study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement. In this negligible risk research study, we analyzed publicly available nonidentifiable aggregated data from the included studies, and our investigation was therefore exempt from formal ethics review. Participants in the included studies did not provide additional consent to that already provided for the primary study.

This review aimed to find, assess, and synthesize all cohort studies that assessed the value of adding a CACS to a standard CVD risk equation (CVD risk), that is, comparing CVD risk plus CACS with CVD risk alone. For a cohort study to be included, it needed to be set in the community or among outpatients without current CVD and use one of the CVD risk calculators recommended by national guidelines (Framingham Risk Score, QRISK, pooled cohort equation [PCE], NZ PREDICT, NORRISK, or SCORE [Systematic Coronary Risk Evaluation) and use cardiovascular events as the outcome for the risk prediction models. The PICO (participants, interventions, comparators, and outcomes) criteria therefore included the following: participants were adults without cardiovascular disease (eg, the general population); interventions consisted of a CACS plus a CVD risk equation; comparators consisted of the same CVD risk equation alone; outcomes consisted of coronary heart disease (CHD) events, stroke events, or a combination; and study designs consisted of cohort studies that reported on subsequent cardiovascular events.

We excluded studies that only reported on a stratum of patients restricted by their CVD risk estimate. Analyses that use restricted CVD risk strata underestimate the discrimination of the CVD risk equations, and hence artifactually overestimate the gain from a CACS (L.Z., K.J.L.B., A.M.S., and P.G., unpublished data, March 2022).

We used a multiple iteration citation search to identify relevant studies for this systematic review. Before the citation search was conducted, a 2018 relevant systematic review by Lin et al¹ was identified. This review was found by searching the website of the USPSTF for the most recent systematic review that evaluated nontraditional risk factors in CVD risk assessment. The USPSTF investigators searched MEDLINE via Ovid, PubMed, and the Cochrane Central Register of Controlled Trials to May 22, 2017. They also assessed included studies and reference lists from 2 previous USPSTF reviews for relevance, as well as multiple other systematic reviews. Because their database search was considered comprehensive, we supplemented their search with a citation search to attempt to identify relevant studies published since their review was published.

We therefore performed an initial backward (cited articles in a reference list) and forward (articles that cite a study) citation search on the systematic review by Lin et al.¹ The citation search, conducted using the Scopus citation database on July 1, 2021, yielded 161 references, which were subsequently screened by title and abstract against the inclusion criteria, and 150 were excluded. The remaining 11 references were primary, potentially includable studies.

The subsequent backward and forward citation searches were performed on these 11 potentially includable studies using the Scopus citation database on July 27, 2021. These searches yielded 2610 references, from which irrelevant references were screened out with a broad topic search on “all fields,” (in EndNote) using the following search string: (Cohort OR [Years AND Follow]) AND (Calcium OR atherosclerosis OR calcification) AND (Score OR Predicted OR Prediction). This search reduced the 2610 references to 457 (ie, excluded 2153), which were then all screened manually.

Articles were screened by title and abstract independently by 2 authors (S.W. and O.H.). Another author (J.C.) retrieved full texts, and the 2 screening authors (S.W. and O.H.) screened the full texts for inclusion. Discrepancies were resolved by a third author (K.J.L.B. or P.G.). The selection process was recorded in sufficient detail to complete a PRISMA flow diagram (eFigure 1 in the Supplement) and a list of excluded (full-text) studies with reasons (eTable 1 in the Supplement). When multiple reports of the same cohort were identified, we selected the most recent that provided data on the most participants.

A standardized form (initially piloted in 2 included studies^13,14) was used for data extraction of the characteristics of studies, outcomes, and risk of bias (eTable 2 in the Supplement). Two authors independently extracted data for each study (K.J.L.B. and either S.W. or O.H.). The study characteristics and outcomes extracted from each included study are provided in eTable 2 in the Supplement. Race and ethnicity data were collected in the primary studies, usually by self-report. These data are relevant to the study because race and ethnicity have been found to be associated with risk of CVD.

Two review authors (K.J.L.B. and A.M.S.) independently assessed the risk of bias for each included study using a modified Quality in Prognosis Studies tool.¹⁵ Each potential source of bias was graded as low, high, or unclear; the addition of the unclear rating was a modification to the original Quality in Prognosis Studies tool. All disagreements were resolved by consensus. The following domains of bias were assessed: (1) study participation, by representativeness of the study sample; (2) study attrition, by participants with follow-up data; (3) prognostic factor measurement, focusing specifically on the CAC scan measurement; (4) outcome measurement (ie, adequacy); (5) study confounding, focusing specifically on the potential for selection bias from the decisions leading to whether participants did or did not receive a CAC scan; and (6) statistical analysis and reporting (ie, appropriateness).

We undertook a meta-analysis of our primary outcome: change in C statistic for the model including the CACS compared with the base model. Study-specific 95% CIs for the change in C statistic were determined by the following hierarchy: (1) the 95% CI for the change in C statistic was reported by the primary study; (2) if a 95% CI was not reported, then the 95% CI was approximated on the basis of the reported exact P value; and (3) if neither a 95% CI nor exact P value were reported, we estimated the 95% CI using Hanley-McNeil method.¹⁶ A detailed explanation of the calculations is provided in eFigure 2 in the Supplement.

Heterogeneity was assessed using Cochran Q test and quantified using the I² statistic. We used random-effects models to pool estimates for each study. Forest plots were used to display the change of C statistic overall and for individual trials. The analyses were performed with R, version 3.6.3 (R Project for Statistical Computing).

The unit of analysis was at the study level (aggregate data). In terms of missing data, we did not receive further data from investigators or study sponsors. Publication bias and small studies effect were not assessed owing to a small number of included studies.

We undertook 3 types of sensitivity analysis. First, we applied the Hanley-McNeil method to all studies as a cross-check on the C statistic estimates derived from the first 2 methods. Second, we estimated gain in C statistic restricted to studies that used published coefficients (models that used published coefficients from a CVD risk equation) rather than deriving coefficients from the study cohort data. Third, we included data from studies that did not meet our inclusion and exclusion criteria but were considered important cohorts for this topic.

As a subgroup analysis, we estimated gain in C statistic for CHD events compared with CVD events. We planned to estimate the gain in C statistic for women compared with men but did not have sufficient data to undertake this analysis. Finally, we had planned an analysis by CVD risk strata, but did not proceed because in preparing the report, we discovered that stratifying the analyses into risk groups yields very biased estimates for incremental gain in C statistic (L.Z., K.J.L.B., A.M.S., and P.G., unpublished data, March 2022). In studies that performed stratified analyses, the C statistic for all risk strata was below the overall mean. This apparent weaker predictive ability is owing to the artificial constriction of the CVD risk predictor model and the nature of the discrimination measure, but this bias is not intuitive. We provide an explanation when and why it happens elsewhere (L.Z., K.J.L.B., A.M.S., and P.G., unpublished data, March 2022).

After reviewing the included studies, we additionally undertook qualitative synthesis of secondary outcomes of incremental prognostic value (hazard ratio of the CACS adjusted for risk equation) and reclassification of participants across treatment thresholds (commonly the 10% 10-year risk) by CACS. We calculated proportions of participants classified as being at low risk or as being at intermediate or high risk by the risk equation who were correctly and incorrectly reclassified when CACS was added to the multivariable model (those at low risk or at intermediate or high risk according to the risk equation used as the denominators). We also calculated absolute rates of the study population who were correctly and incorrectly reclassified (total study sample used as the denominator).¹⁷ Correct or incorrect reclassification was determined according to whether or not the person had a CVD event during follow-up.

In total, 2772 records were identified through searching other sources: 161 records from the citation search performed by Lin et al,¹ 2610 from the second citation search, and 1 additional record from the reference list of a review found by the citation search. The title and abstract were screened for 161 records from the initial citation search, 457 records (of 2610) from the second citation search (see “Search Strategy” section in Methods), and the 1 record from the reference list, resulting in 53 full texts to screen, of which 47 were excluded (for reasons, see eFigure 1 and eTable 1 in the Supplement). Six studies^13,14,18-21 were included in the qualitative synthesis and in the meta-analyses (eFigure 1). Rates of cardiovascular events ranged from 22 of 2303 (1.0%)²¹ to 347 of 3678 (9.4%).¹⁹

The 6 included studies (with 1043 CVD events in 17 961 unique participants) were from the US (n = 3),^14,18,21 the Netherlands (n = 1),¹⁹ Germany (n = 1),¹³ and South Korea (n = 1)²⁰ and ranged in size from 470 to 5185 participants (Table 1). The mean (SD) age of study participants ranged from 50 (10) to 75.1 (7.3) years; 38.4% to 59.4% were women; 40.6% to 61.6% were men; and 38% to 100% were White. Predictive factors included in the base model were published coefficients from the Framingham Risk Score 1998 (n = 2), PCE (n = 2), or derived coefficients for risk factors (n = 2)

For the study participation domain, most of the included studies were at low risk of bias,^13,14,18-20 although 1 of the 6 studies²¹was rated as being at high risk of bias in this domain owing to nonrepresentativeness of the study sample (Figure 1 and eFigure 3 in the Supplement). Most studies were rated as being at unclear risk of bias for attrition,^13,14,18,19 although 2 were rated as being at high risk of bias^20,21 owing to considerable attrition from the parent cohort. Three studies were rated as being at low^14,18,20 and 3 at unclear^13,19,21 risk of bias for prognostic factor measurement. Five studies were rated as being at low risk of bias for outcome measurement,^{13,14,18,19,21} and 1 was rated as being at unclear risk.²⁰ One study¹³ was rated as being at low risk owing to confounding; the remainder^14,18-21 were rated as being at unclear risk, because it was often unclear which participants did—and which did not—receive the CAC scan, and the reasons for those decisions. For statistical analysis and reporting, 5 studies^13,14,18-20 were rated as being at low risk of bias; 1 study²¹ was rated as being at high risk of bias.

The C statistic for the CVD risk models without CACS ranged from 0.693 (95% CI, 0.661-0.726) to 0.80 (Table 2). The pooled gain in C statistic was 0.036 (95% CI, 0.020-0.052) with moderate heterogeneity (I² = 56%) (Figure 2). The heterogeneity may be partly explained by the outcome measures: the gain appeared larger for studies using CHD events as the outcome (gain in C statistic of 0.049 [95% CI, 0.036-0.062]; I² = 0) than for studies using CVD events (gain in C statistic, 0.029 [95% CI, 0.011-0.047]; I² = 39%). However, the test for interaction indicated that this difference was not statistically significant (P = .08).

For our sensitivity analyses, we conducted 2 additional analyses. Applying the Hanley-McNeil method to all studies did not change the gain in C statistic (0.037 [95% CI, 0.021-0.052]) or the apparent difference for studies using CHD events as the outcome (gain in C statistic of 0.049 [95% CI, 0.036-0.062]) and studies using CVD events (gain in C statistic of 0.032 [95% CI, 0.014-0.050]) (eFigure 4 in the Supplement). Restricting the meta-analysis only to the 4 studies that used a published risk equation^13,14,20,21 yielded a comparable gain in C statistic to the pooled estimate from all 6 studies (0.034 [95% CI, 0.012-0.056]). Including the Coronary Artery Calcium Consortium Study²² in the meta-analysis also yielded a similar gain in C statistic (0.034 [95% CI, 0.022-0.046] after including the result of CHD mortality; 0.032 [95% CI, 0.018-0.046] after including the result of CVD mortality) (eFigure 4 in the Supplement).

Adjusted hazard ratios for the CACS ranged from 1.29 per SD (1.58 per 1.98 SD; 95% CI, 1.40-1.79) to 1.75 per SD (95% CI, 1.37-2.44). Four studies^13,14,18,19 reported reclassification tables that allowed calculation of the proportion and absolute rates of correct and incorrect reclassification when the CACS was added to the risk model (Geisel et al¹³ included additional data from Erbel et al²³) (Table 3). Among participants classified as being at low risk by the risk equation, 0.4% (11 of 3139)¹⁸ to 2.2% (54 of 2471)¹⁹ were correctly reclassified as being at intermediate or high risk when a CACS was added to the model (ie, had a CVD event), and 2.1% (65 of 3139)¹⁸ to 14.4% (349 of 2416)¹³ were incorrectly reclassified as being at intermediate or high risk (did not have an event). The absolute rates in the study populations (total study sample used as the denominator) were 0.3% (13 of 4129)¹³ to 1.5% (54 of 3678)¹⁹ correctly reclassified, and 2.0% (65 of 3319)¹⁸ to 9.6% (496 of 5185)¹⁴ incorrectly reclassified. Among participants reclassified from low risk by the risk score to intermediate or high risk by CACS, 3.6% (13 of 362)¹³ to 14.5% (11 of 76)¹⁸ had a CVD event during follow-up, and 85.5% (65 of 76)¹⁸ to 96.4% (349 of 362)¹³ did not have a CVD event during follow-up.

Among participants classified as being at intermediate or high risk by the risk equation, 18.9% (34 of 180)¹⁸ to 29.3% (502 of 1713)¹³ were correctly reclassified as being at low risk when CACS was added to the model (did not have a CVD event), and 0.2% (4 of 1713)¹³ to 1.9% (19 of 1000)¹⁴ were incorrectly reclassified as being at low risk (had a CVD event). The absolute rates in the study populations were 1.0% (34 of 3319)¹⁸ to 12.2% (502 of 4129)¹³ correctly reclassified as being at low risk and 0.1% (2 of 3319)¹⁸ to 0.5% (19 of 3678)¹⁹ incorrectly reclassified as being at low risk. Among participants reclassified from intermediate or high risk by the risk score to low risk by CACS, 91.4% (202 of 221)¹⁴ to 99.2% (502 of 506)¹³ did not have a CVD event during follow-up, and 8.6% (19 of 221)¹⁴ to 0.8% (4 of 506)¹³ did have a CVD event during follow-up.

We found 6 cohort studies^13,14,18-21 that could provide an estimate of the gain in discrimination of a CACS beyond the traditional CVD risk factors included in CVD risk equations. Despite variation in the overall risk of the study populations (event rates ranging from 1.0% to 9.4%), the improved discrimination was relatively consistent. The mean discrimination gain of 0.036 is modest, but may be important in some subgroups, and adjusted hazard ratios suggest the test may have incremental prognostic value for some participants (eTable 3 in the Supplement). However, 85.5% to 96.4% of participants reclassified as being at intermediate or high risk by CACS did not have a CVD event during follow-up. On the other hand, CACS may have clinical utility in downgrading risk, because 91.4% to 99.2% of cases reclassified as being at low risk by CACS also did not have a CVD event during follow-up, but it would be important to identify the nontrivial 0.8% to 8.6% of participants incorrectly reclassified as being at low risk by CACS to ensure these individuals did not miss the opportunity for beneficial preventive treatment. Also, our analysis assumed that a CACS was added to the multivariable risk prediction model to arrive at new risk estimates, but in clinical practice the CACS is used separately from the CVD risk equation, and the clinical utility may be lower.

The 2018 USPSTF systematic review¹ included 18 publications evaluating discrimination with the CACS in participants from a broad range of primary prevention populations using the Framingham Risk Score or the PCE. However, the USPSTF review included several studies that used the same cohorts and studies that were restricted to CVD risk strata, and it did not attempt to combine the studies. Despite these limitations, the investigators’ overall conclusions were similar to ours. Our meta-analysis showing modest gain in discrimination and our reclassification analysis exploring potential clinical utility add to the existing understanding of the use of a CACS for CVD assessment. Our search for studies published after the USPSTF evidence review identified 4 potentially relevant studies. However, 2 performed further analysis of the Multi-Ethnic Study of Atherosclerosis cohort: one was a substudy of patients at intermediate risk only²⁴ and the other was part of an already published 3-cohort analysis²⁵ (eTable 1 in the Supplement). A pooled analysis by the Coronary Artery Calcium Consortium²² only provided data on CVD deaths, not events, and so was excluded, but the gain in C statistic was consistent with our finding: a gain in C statistic over the PCE of 0.02 for CVD deaths and 0.03 for CHD deaths. Sensitivity analyses including these data yielded similar meta-analytic estimates for the gain in discrimination (eFigure 4 in the Supplement). An analysis of the Walter Reed Cohort Study²⁶ adjusted for baseline risk factors but not covariates of CVD risk calculators, and so was excluded; however, the gain in C statistic was again consistent at 0.02 for CVD and 0.08 for CHD. Hence, the only new data were from a small study in South Korea (Korean Longitudinal Study on Health and Aging).²⁰

The CACS appears to add some further discrimination to the traditional CVD risk assessment equations used in these studies, and the effect appears relatively consistent across studies. The gain is modest and would need to be weighed against the costs and radiation risks. Further, if a CACS is used to reclassify individuals as being at high risk but not to reclassify individuals as at low risk, then this will expand the number of people in the high-risk category. Our finding that most participants reclassified as being at high risk did not have a CVD event during follow-up suggests use of a CACS has potential to cause harm from inappropriate diagnostic labeling, unnecessary treatments, and unnecessary tests. The substantial reservoir of subclinical disease (present in 42% of adults without known CVD, of which 88% is classified as nonobstructive)²⁷ presents a large potential for overdiagnosis if a CACS is used indiscriminately in the population.²⁸ Further, CAC density is inversely associated with CVD risk and may actually indicate stabilization of atherosclerotic plaque.²⁹

Although the overall improvement in CVD risk discrimination from a CACS is modest—as shown by the gain in C statistics—it may have a role for refining risk assessment in selected patients.³⁰ This possible role is suggested by the hazard ratios for the CACS, which ranged from 1.29 to 1.75 per SD, after adjustment for other CVD risk factors. The groups most likely to benefit are patients for whom, after standard CVD risk score assessment, reasonable likelihood exists that CACS could help “in clarifying whether the risk is high enough to justify primary prevention medications.”³⁰ However, the gains for most such subgroups of patients are unclear at present, and new methods are needed to assess the incremental gains in such subgroups. Further refinement is likely to need individual patient data to define who may benefit from a CACS, but methods are needed that avoid using strata defined by CVD risk equations to assess the incremental gain, because such stratified analysis is seriously flawed (L.Z., K.J.L.B., A.M.S., and P.G., unpublished data, March 2022). Although providing disaggregated results by sex and other important subgroupings should be encouraged in primary publication,³¹ contribution of individual level data will be important for refining such subgroups.

Despite a thorough search, risk of bias assessment, and pooled analysis, our review has some possible limitations. First, the attrition of participants from the studies included indicated unclear or high risk of bias. However, this attrition applies to measurement of both the CACS and traditional CVD risk factors and therefore may not have biased estimates of incremental value. Second, the studies only used the Framingham Risk Score and PCE CVD risk equations to evaluate potential gain from a CACS, and the incremental gain may be smaller for other risk equations that include more risk factors, such as the QRISK and PREDICT equations.^32,33

This systematic review and meta-analysis suggests that the CACS adds further discrimination to the traditional CVD risk assessment equations used in these studies, and the effect appears relatively consistent across studies. However, the modest gain may often be outweighed by costs, rates of incidental findings, and radiation risks. Although a CACS may have a role for refining risk assessment in selected patients, which patients would benefit remains unclear. At present, no evidence suggests that adding the CACS to traditional risk scores provides clinical benefit.

Accepted for Publication: March 5, 2022.

Published Online: April 25, 2022. doi:10.1001/jamainternmed.2022.1262

Correction: This article was corrected on August 1, 2022, to correct data errors that appeared in the Results, the Discussion, Table 2, Table 3, and eTable 3. This article was corrected on June 6, 2022, to correct 2 erroneous citations of references 30 and 31 in the Discussion.

Corresponding Author: Katy J. L. Bell, PhD, School of Public Health, University of Sydney, Edward Ford Building (Building A27), Sydney, NSW 2006, Australia (katy.bell@sydney.edu.au).

Author Contributions: Drs Bell and Zhu had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Bell, White, Scott, Glasziou.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Bell, Zhu, Scott, Glasziou.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Bell, White, Hassan, Zhu, Glasziou.

Administrative, technical, or material support: White, Hassan, Scott, Clark.

Supervision: Bell, Glasziou.

Conflict of Interest Disclosures: Dr Bell reported receiving grants from the Australian National Health and Medical Research Council (NHMRC) and salary and project support from an Investigator grant during the conduct of the study. Dr Scott reported receiving grants from National Heart Foundation of Australia during the conduct of the study. Mr Clark reported receiving grants from the National Heart Foundation of Australia during the conduct of the study. Dr Glasziou reported receiving grants from the NHMRC and the National Heart Foundation of Australia during the conduct of the study. No other disclosures were reported.

Funding/Support: This study was supported by NHMRC Investigator grant 1174523 (Dr Bell) and NHMRC Australian Fellowship grant 108004 (Dr Glasziou).

Role of the Funder/Sponsor: The National Heart Foundation of Australia, who commissioned this study as part of a series of systematic reviews to support the development of guidelines for absolute cardiovascular disease risk, and the NHMRC had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.