A Systematic Review and Meta-Analysis of Advanced Biomarkers for Predicting Incident Cardiovascular Disease among Asymptomatic Middle-Aged Adults

Cardiovascular disease (CVD) continues as the most important cause of mortality. Better risk screening and prediction are needed to reduce the cardiovascular disease burden. The aim of the study was to assess the role of serum biomarkers in the prediction of CVD among asymptomatic middle-aged adults with no prior CVD history. A systematic review and meta-analysis were carried out using literature from PubMed and following PRISMA reporting guidelines. Twenty-five studies met our inclusion criteria and were included in the systematic review. The most commonly studied biomarker was high-sensitivity C reactive protein (hs-CRP) (10 studies), which showed that higher hs-CRP levels are associated with an increased risk of subsequent CVD events and mortality. In addition, several less-studied biomarkers (N-terminal pro-brain natriuretic peptide (NT-proBNP), fibrinogen, gamma-glutamyl transferase (GGT), and others) also showed significant associations with greater future risk of CVD. A meta-analysis was possible to perform for hs-CRP and NT-proBNP, which showed statistically significant results for the ability of hs-CRP (hazard ratio (HR) 1.19, (95% CI: 1.09–1.30), p < 0.05) and NT-proBNP (HR 1.22, (1.13–1.32), p < 0.05) to predict incident CVD among middle-aged adults without a prior CVD history or symptoms. Several serum biomarkers, particularly hs-CRP and NT-proBNP, have the potential to improve primary CVD risk prevention among asymptomatic middle-aged adults.


Introduction
Despite concerted efforts from clinicians, researchers and public health authorities to prevent and reduce the burden of cardiovascular disease (CVD), CVD remains a leading cause of morbidity and mortality worldwide [1].
The evidence base for controlling traditional CVD risk factors, such as lipids, blood pressure and blood glucose along with smoking cessation, to prevent and reduce CVD events is well-established, and clinical and population health measures in this regard have improved over the last decades. However, given the ongoing and large burden of CVD morbidity and mortality, significant research has been directed to studying additional tools, such as novel biomarkers and imaging techniques, that may prove to be better predictors of disease outcomes or permit earlier detection of CVD. Assessment of cardiovascular disease risk using biomarker analysis holds promise as a more effective approach for stratifying those at high risk for CVD and for optimizing the prevention and treatment of

Search Strategy and Study Selection Criteria
The review protocol was registered in PROSPERO, the International Prospective Register of Systematic Reviews (crd.york.ac.uk/prospero/index.asp, identifier: CRD42020167521). An advanced search in the PubMed/Medline database with specific keywords related to the topic of interest was conducted using MeSH terms to create a uniform search strategy using some biomarkers with previous evidence (fibrinogen, hs-CRP, myeloperoxidase, homocysteine, IL-6, lipoprotein (a), apolipoproteins, lipoprotein-associated phospholipase A2, oxidized LDL or metabolomics) and cardiovascular outcomes (coronary heart disease, heart failure, stroke, peripheral artery disease or CVD mortality). In addition to filters for language, the age or date of publication was used (full details in Supplementary Table S1). Data collection was performed from December 2019 to February 2020, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guidelines [12]. Search parameters were limited to the following filters: language (English), age (over 18 years and through 65 years), articles published in the last 10 years (November 2009-November 2019), and the types of studies included were cross-sectional studies and case-control and cohort studies. Participants in the selected studies had no known history of cardiovascular disease (coronary heart disease, heart failure, stroke or peripheral artery disease). A flow diagram summarizes the process of study selection (Figure 1).

Data Extraction and Quality Assessment
After the initial literature search, duplicate studies, conference abstracts, editorials, and letters to the editor were removed, and two investigators (JLRC and AFM) independently screened article titles and abstracts and decided if each study was eligible for further review. Inter-rater agreement was assessed, and discrepancies between the two indepen-dent primary reviewers were resolved by one of the senior investigators (SNK or DS). The inclusion criteria were as above: age (between 18 and 65 years), no prior or current history of cardiovascular disease, types of studies (case-control and cohort studies), and emerging/novel/advanced biomarkers beyond glucose and classic lipid profiles. The exclusion criteria were age (participants older than 65 years), current or prior history of cardiovascular disease, or other types of studies such as clinical trials, case reports or meta-analyses. After the screening of article titles and abstracts, 2055 articles were removed because they did not meet the inclusion criteria. the selected studies had no known history of cardiovascular disease (coronary heart disease, heart failure, stroke or peripheral artery disease). A flow diagram summarizes the process of study selection ( Figure 1).

Figure 1.
Flow diagram illustrating the search strategy used according to PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement.

Data Extraction and Quality Assessment
After the initial literature search, duplicate studies, conference abstracts, editorials, and letters to the editor were removed, and two investigators (JLRC and AFM) independently screened article titles and abstracts and decided if each study was eligible for further review. Inter-rater agreement was assessed, and discrepancies between the two independent primary reviewers were resolved by one of the senior investigators (SNK or After reading the full text and supplementary materials of each study, the data extraction form captured the following items: publication details, study participants, type and description of the biomarkers, type of cardiovascular disease outcomes, and author conclusions. To assess the validity of the studies, we reviewed and evaluated the quality of each study on the basis of the criteria created by the third USPSTF (US Preventive Services Task Force) [13].

Data Synthesis and Analysis
We carried out the meta-analysis using meta command in Stata 15.0 (Stata Corp LP, College Station, TX, USA) to compute summary results for incident CVD outcome risk from eligible studies for each biomarker with sufficient individual results (at least three studies) for middle-aged adults without a prior CVD history or symptoms. The primary study outcomes were incident CVD, defined as any of the following: myocardial infarction, coronary revascularization procedures, stroke, heart failure, cardiac arrest, peripheral artery disease, or death from CVD. We used adjusted HRs to pool results when at least three studies were found for a given variable with comparable methodologies and results. The heterogeneity between the results of the studies was examined, and the magnitude of heterogeneity was judged by combining the findings with the Cochran's Q test and the I 2 test with values 25%, 50%, and 75% representing low, moderate, and high heterogeneity, respectively [14]. A fixed-effects model or a random-effects model was used as appropriate with 95% confidence intervals. To examine publication bias in the included research, an Egger's test was performed, with values p > 0.10 considered no risk of bias.

Results
We identified 2264 studies through our search strategy. Based on a review of the titles and abstracts, 2055 were removed. After reading the full text of all articles, 183 studies were removed because their participants were older than 65 years old or had a previous history of CVD. Ultimately, 25 studies met all inclusion criteria with a total sample size of 287,020 participants. These included 20 prospective cohort studies, 3 nested case-control studies and 2 retrospective case-control studies . The summary of the 25 studies included is shown in Supplementary Table S2 (Supplementary Material). The quality of most of the studies was good. The biomarker mostly commonly reported in these studies was hs-CRP (10 studies), NT-proBNP (4), fibrinogen (3), GGT (3), homocysteine (2), hsTn (2), and suPAR (2). Separated tables summarize the results for each study, such as type of study, quality of the study, population, CV outcome definitions, length of follow-up and main results for those biomarkers with at least three studies (Tables 1 and 2,  and Supplementary Tables S3 and S4). In addition, it was possible to perform a metaanalysis to calculate the joint results of the hs-CRP and NT-proBNP studies predicting incident CVD among middle-aged adults without a prior CVD history or symptoms. The confounder factors that were mostly used for adjustment were age, gender, race and smoking status or traditional cardiovascular risk factors such as hypertension, dyslipidemia and diabetes mellitus.

Cohort (Prospective study) Good
A total of 6621 participants aged 45-84 years from the MESA study (these participants were excluded from the current analysis) and 2202 participants aged 30-65 years from the DHS study.

Cohort (Prospective study) Good
A total of 15,737 participants from the Scottish Heart Health Extended Cohort (SHHEC) with a mean age of 49 (SD 8.3) years.
A mean follow-up period of 19.9 years.

Cohort (Prospective study) Good
A total of 6974 participants from the PREVEND cohort were included in the current study with a mean age of 48 years old.
A median follow-up period of 10.5 years.
A total of 737 incident CVD events were recorded, and the association of hs-CRP with incident CVD was shown. The analysis showed an HR of 1 Participants were followed for a mean of 10 years for the occurrence of a first major cardiovascular event (CVE).
The total cohort was divided into hormone non-users and hormone users. Firstly, in a continuous analysis, the RR d for lnCRP was 1.27 (95%: CI, 1.13 to 1.44) for HT e non-users and 1.22 (95%: CI, 1.07 to 1.40) for HT users. In addition, the relative risk was calculated according to the quintile of CRP based on HT non-users and HT users levels and categories defined by AHA/CDC (<1 mg/L, ≥1-<3 mg/L, and ≥3 mg/L). After risk factor-adjusted RR, the highest quintile, based on HT non-users levels (>4.18 mg/L), significantly predicted CVE in HT non-users RR: 2.85 (95% CI: 1.62 to 5.00), but not in HT users. However, based on HT users levels (>6.44 mg/L), the highest quintile predicted CVE RR: 1.88 (95% CI: 1.14 to 3.11). In a fit model in which non-users with CRP< 1 mg/L were the reference group after adjusting HT users with CRP ≥3 had a RR of 1   [39].

Cohort (Prospective study) Fair
A total of 1988 healthy subjects were included after the exclusion of 472 subjects from the baseline study with known diabetes, prior myocardial infarction or stroke. These healthy subjects were classified using HeartScore as high risk (559) and low-moderate risk (1429), depending on the expected 10-year risk of CV death above or below 5%.
Cardiovascular death, non-fatal MI or stroke.
A follow-up period of 9.5 years.

Cohort (Prospective study) Fair
A total of 1988 healthy subjects were included after the exclusion of 472 subjects from the baseline study with known diabetes, prior myocardial infarction or stroke. These healthy subjects were classified using HeartScore as high risk (559) and low-moderate risk (1429), depending on the expected 10-year risk of CV death above or below 5%.
Cardiovascular death, non-fatal myocardial infarction or stroke.
A follow-up period of 9.5 years.
The ability of hs-CRP to predict incident CVD among middle-aged adults without a prior CVD history or symptoms was computed by performing a meta-analysis. The metaanalysis included seven of the ten studies that were found with comparable continuous analysis, using an adjusted hazard ratio per increment of 1 SD unit of the continuous predictor variable (Figure 2). Separated results were provided in one study because joint results were not available. High heterogenicity between studies (p < 0.01, Q = 47.59, I 2 = 85.3%) necessitated the use of the random-effects model for meta-analysis. Despite high heterogenicity, statistically significant results for the ability of hs-CRP to predict incident CVD were found (HR 1.17, (95%: CI: 1.06-1.30), p < 0.05). Egger's test was performed to assess the publication bias and showed significant results (p = 0.03). Therefore, sensitivity analyses were performed. The study by Zöga Diederichsen et al. (2018) [17], the study with the shortest length of follow-up, was deleted, showing lower heterogenicity (p < 0.01, Q = 18.01, I 2 = 66.7%) and similar results to predict incident CVD (HR 1.21, (95% CI: 1.09-1.33), p < 0.05), in addition, the Egger's test performed was not significant (p = 0.37) (Figure 3).
A total of 231,393 participants were included in the 10 studies that assessed the association between hs-CRP and cardiovascular outcomes among middle-aged, asymptomatic individuals (Table 1) [16][17][18][19][20]26,34,35,38,39]. All 10 studies were prospective studies. Generally, hs-CRP was found to be related to a higher risk of CVD and CV mortality.
The ability of hs-CRP to predict incident CVD among middle-aged adults without a prior CVD history or symptoms was computed by performing a meta-analysis. The metaanalysis included seven of the ten studies that were found with comparable continuous analysis, using an adjusted hazard ratio per increment of 1 SD unit of the continuous predictor variable (Figure 2). Separated results were provided in one study because joint results were not available. High heterogenicity between studies (p < 0.01, Q = 47.59, I 2 = 85.3%) necessitated the use of the random-effects model for meta-analysis. Despite high heterogenicity, statistically significant results for the ability of hs-CRP to predict incident CVD were found (HR 1.17, (95%: CI: 1.06-1.30), p < 0.05). Egger's test was performed to assess the publication bias and showed significant results (p = 0.03). Therefore, sensitivity analyses were performed. The study by Zöga Diederichsen et al. (2018) [17], the study with the shortest length of follow-up, was deleted, showing lower heterogenicity (p < 0.01, Q = 18.01, I 2 = 66.7%) and similar results to predict incident CVD (HR 1.21, (95%: CI: 1.09-1.33), p < 0.05), in addition, the Egger's test performed was not significant (p = 0.37) ( Figure  3).

N-Terminal Pro-Brain Natriuretic Peptide (NT-proBNP)
Four studies, including a total of 24,055 participants, were assessed for NT-Pro-BNP and risk of CVD and mortality (Table 2) [19,20,29,39]. These four studies were prospective studies, and most of them found that NT-Pro-BNP was associated with CVD among our population.
For NT-proBNP, it was possible to perform a meta-analysis including the four studies with continuous analysis to assess the association of NT-proBNP level with CVD incidence, using an adjusted hazard ratio per increment of 1 SD unit of the continuous predictor variable (Figure 4). A random-effects model for meta-analysis was needed because there was high heterogeneity between studies (p = 0.052, Q = 9.4, I 2 = 57.5%). Despite high heterogenicity, the results showed the statistically significant ability of NT-proBNP to predict incident CVD among middle-aged adults without a prior CVD history or symptoms (HR 1.22, (1.13-1.32), p < 0.05). Egger's test was performed to assess the publication bias, and it was not significant (p = 0.38).  Figure 3. Forest plot of sensitivity analysis of studies for hs-CRP prediction of CVD among middleaged adults without a prior CVD history or symptoms [16,19,20,26,38,39].

N-Terminal Pro-Brain Natriuretic Peptide (NT-proBNP)
Four studies, including a total of 24,055 participants, were assessed for NT-Pro-BNP and risk of CVD and mortality (Table 2) [19,20,29,39]. These four studies were prospective studies, and most of them found that NT-Pro-BNP was associated with CVD among our population.
For NT-proBNP, it was possible to perform a meta-analysis including the four studies with continuous analysis to assess the association of NT-proBNP level with CVD incidence, using an adjusted hazard ratio per increment of 1 SD unit of the continuous predictor variable (Figure 4). A random-effects model for meta-analysis was needed because there was high heterogeneity between studies (p = 0.052, Q = 9.4, I 2 = 57.5%). Despite high heterogenicity, the results showed the statistically significant ability of NT-proBNP to predict incident CVD among middle-aged adults without a prior CVD history or symptoms (HR 1.22, (1.13-1.32), p < 0.05). Egger's test was performed to assess the publication bias, and it was not significant (p = 0.38).  [16,19,20,26,38,39].

Fibrinogen
Three prospective studies were identified that assessed the risk of elevated fibrinogen levels for CVD (Supplementary Table S3) [20,31,37]. Two studies showed a significantly higher risk for CVD with higher fibrinogen levels using continuous statistical analysis, with an HR of 1.09 (CI 95%: 1.05-1.12) in one study and an HR of 1.09 (CI 95%: 1.02-1.19) for men and 1.09 (CI 95%: 1.01-1.18) for women in the other study [20,37]. However, another study among three central Australian Aboriginal communities found a non-significant association of fibrinogen with CVD events when using a cut-off of ≥3.5 g/L and an OR of 1.62 (CI 95%: 0.95, 2.78) [31].

Gamma-Glutamyl Transferase (GGT)
Three prospective studies assessed the risk of higher GGT levels and incident CVD events [20,27,31] (Supplementary Table S4). GGT was associated with a higher risk in all three studies, two of which used continuous analysis with an HR of 1.07 (CI 95%: 1.02-1.13) and 1.18 (CI 95%: 1.06-1.30), respectively, and one used a cut-off of ≥70 U/L with an OR of 2.66 (CI 95%: 1.56-4.55) [31]. Subanalysis was carried out in one of the studies to assess the risk of stroke and CHD separately [27]. This study found that high GGT was a significant risk factor for stroke, with an HR of 1.38 (CI 95%: 1.12-1.69), but not for CHD, with an HR of 1.09 (CI 95%: 0.95-1.26).

High-Sensitive Troponin (hsTn)
Only two cohort prospective studies were found that assessed the association between hsTn levels and CVD among asymptomatic middle-aged individuals [19,20]. Each of these studies showed a significant risk for CVD with higher hsTn levels. In the prospective cohort study by de Lemos et al. (2017) [19], both categorical (≥5 µg/L or highest quartile) and continuous analyses indicated that higher hsTn levels were associated with higher CVD risk: HR of 1.46 (CI 95%: 1.01-2.11) and HR of 1.17 (CI 95%: 1.01-1.17), respectively.

Fibrinogen
Three prospective studies were identified that assessed the risk of elevated fibrinogen levels for CVD (Supplementary Table S3) [20,31,37]. Two studies showed a significantly higher risk for CVD with higher fibrinogen levels using continuous statistical analysis, with an HR of 1.09 (CI 95%: 1.05-1.12) in one study and an HR of 1.09 (CI 95%: 1.02-1.19) for men and 1.09 (CI 95%: 1.01-1.18) for women in the other study [20,37]. However, another study among three central Australian Aboriginal communities found a non-significant association of fibrinogen with CVD events when using a cut-off of ≥ 3.5 g/L and an OR of 1.62 (CI 95%: 0.95, 2.78) [31].

Gamma-Glutamyl Transferase (GGT)
Three prospective studies assessed the risk of higher GGT levels and incident CVD events [20,27,31] (Supplementary Table S4). GGT was associated with a higher risk in all three studies, two of which used continuous analysis with an HR of 1.07 (CI 95%: 1.02-1.13) and 1.18 (CI 95%: 1.06-1.30), respectively, and one used a cut-off of ≥ 70 U/L with an OR of 2.66 (CI 95%: 1.56-4.55) [31]. Subanalysis was carried out in one of the studies to assess the risk of stroke and CHD separately [27]. This study found that high GGT was a significant risk factor for stroke, with an HR of 1.38 (CI 95%: 1.12-1.69), but not for CHD, with an HR of 1.09 (CI 95%: 0.95-1.26).

High-Sensitive Troponin (hsTn)
Only two cohort prospective studies were found that assessed the association between hsTn levels and CVD among asymptomatic middle-aged individuals [19,20]. Each

Homocysteine
Only one prospective study and one case-control study were identified that assessed the association of higher homocysteine levels with CVD incidents in middle-aged, asymptomatic individuals [20,30]. In the prospective study, homocysteine analyzed using a continuous statistical analysis showed an HR of 1.08 (CI 95%: 1.05-1.12) for CHD [20]. The case-control study included participants with CVD and Metabolic Syndrome (MS) and found that homocysteine levels were higher in MS and CVD cases as compared to normal non-CVD controls. Linear regression analysis using homocysteine confirmed a significant positive correlation for CVD in both non-MS (β = 0.041, p = 0.001) and MS subjects (β = 0.027, p = 0.034).

Soluble Urokinase Plasminogen Activator Receptor (suPAR)
Two prospective studies met the inclusion criteria for the current systematic review and assessed whether higher suPAR levels could predict CVD in middle-aged asymptomatic individuals [17,35]. Eugen-Olsen J. et al. (2010) [35] stratified by age and showed a significantly higher risk for CVD (CV death, ischemic heart disease and stroke) with elevated suPAR in younger subjects (41-and 51-year-old subjects, HRs of 1.32 and 1.22, respectively) and the oldest subjects (71-year-old subjects, HR of 1.18) but not in 61-year-old subjects in the continuous analysis. In addition, the authors analyzed the risk of mortality from any cause based on suPAR and found elevated risk in all age groups (41- Diederichsen and colleagues found that older participants (60 years old) with elevated suPAR had an increased risk of CVD events, but younger participants (50 years old) did not. Diederichsen et al. also performed an analysis with stratification by gender that showed a significantly higher risk between high levels of suPAR and CVD in females (HR 1.96, (CI 95%: 1.41-2.71)) but not in males (HR 1.05, (CI 95%: 0.80-1.36)).

Others
Many studies examined other biomarkers or ratios and met the inclusion criteria based on age and absence of CVD, including articles on uric acid, endostatin, HOMA-IR, lipoprotein (a), cystatin C, AST, ALT, alkaline phosphatase, albumin/creatinine ratio, etc., and their results are shown in the summary table (Supplementary Material) as there were not enough studies on each variable to consider separately.

Discussion
The results of this systematic review and meta-analysis provide important evidence about the relationship between various biomarkers and the risk for CVD events and CVD mortality among middle-aged asymptomatic subjects, especially for hs-CRP and NT-proBNP. Our findings are consistent with previous reviews that investigated the associations between biomarkers and CVD risk in individuals with a previous history of CVD or symptomatic subjects [40] and suggest that the use of biomarkers among middle-aged asymptomatic individuals may improve the ability to detect early atherosclerotic disease.
Atherosclerosis is considered a chronic, systemic, low-grade inflammatory disease of the arterial walls driving a gradual accumulation of lipoproteins from an early age. Initially, atherosclerosis is subclinical and often goes undetected. However, advancing atherosclerosis can lead to unstable atherosclerosis plaque and fatal cardiovascular events in older age [41].
Traditional risk factors, including a family history of premature ASCVD, smoking, age, primary hypercholesterolemia, blood glucose and high triglycerides, have been the foundation of risk identification for several decades. Currently, many CV risk assessment systems, including classical risk factors, are used to estimate the risk of a first fatal CV event in apparently healthy subjects, usually expressed as a 10-year. Examples include the Framingham CVD risk score [42], SCORE (Systematic COronary Risk Evaluation) [43] and ASCVD risk estimator [2]. However, these tools may underestimate risk in many cases. Recent evidence has demonstrated that adding another factor, such as imaging tools, serum biomarkers and different clinical risk factors, could improve the CV risk assessment. Thus, some new CV risk assessment systems have added hs-CRP and/or coronary artery calcification (such as MESA; Multi-Ethnic Study of Atherosclerosis and Astro-CHARM; Astronaut Cardiovascular Health and Risk Modification) to improve the accuracy of the estimation of risk [16,44]. Based on a systematic review of current evidence, the US Preventive Services Task Force concluded there was insufficient evidence to recommend high-sensitivity CRP in overall risk assessment in the population [45]. However, in many cases, particularly among intermediate-moderate risk individuals, biomarkers or other non-traditional risk factors may be useful to make clinical decisions regarding therapeutic measures. Therefore, biomarkers have been widely suggested as a potential tool to help in CV risk assessment. In particular, they may be an important way to better assess risk at the early stages of disease progression and permit the initiation of more aggressive prevention measures, such as changes in lifestyle. This approach may permit more definitive steps at the subclinical stages of disease and stop the development of unstable atherosclerosis plaque and fatal events.
Despite the apparent value of using biomarkers, these biomarkers in the overall population have limitations because they are not specific to the assessment of cardiovascular disease. For example, any other systematic disease, such as rheumatic diseases or infectious diseases in the acute phase, could raise inflammation biomarkers. Therefore, it is necessary to consider biomarkers in the context of multiple different situations [46].
The current American and European guidelines on the primary prevention of cardiovascular disease and for the management of dyslipidemias recently recommended using hs-CRP levels as additional data to enhance CVD risk assessment to inform clinician-patient risk discussions [2,3,43]. Inflammation is clearly in the causal pathway of atherosclerosis, and hs-CRP has been considered one of the most useful and readily available serum markers of inflammation. Thus, there have been suggestions that treatment with high-intensity statins could reduce the hs-CRP levels, inflammation and CV risk. Incorporating hs-CRP into treatment decisions is supported by results obtained in the JUPITER trial [47] and the CANTOS trial [48]. The association between pathophysiological inflammation pathways and atherosclerosis and CVD has also been reported in other studies. Recently, Ronit et al. have shown a robust association between plasma albumin and CVD outcomes and periodontal disease [49,50]. Dietary intake of several substances, such as red meat, alcohol, olive oil, and dairy, has also been shown to either increase or decrease inflammation and the incidence of CVD [51][52][53].
Research continues on other advanced biomarkers that could improve risk prediction for those who might be at high risk but are not identified with traditional risk factor screening. Cardiac biomarkers, such as hsTn and NT-proBNP, are related to higher rates of CVD events and mortality among individuals with advanced age and a history of CVD. In addition, general population cohort studies of older individuals have also demonstrated that these biomarkers are related to higher rates of CVD events and mortality, as well as in a meta-analysis among people without baseline cardiovascular disease. Thus, hsTn and NT-proBNP could improve prediction for CVD among individuals with advanced age and a history of CVD [54]. However, little work has focused on middle-aged subjects younger than 65 years old. Based on the small number of studies that we were able to identify, our findings suggest that higher hsTn and NT-proBNP levels are predictive of higher risk for CVD events and mortality in young, asymptomatic individuals and could be a useful tool for predicting CVD in this group.
Many of the biomarkers that were reported in the literature that we retrieved demonstrated elevated an HR for CVD outcomes, however, the HRs were modest, and it is often unclear how much predictive ability they add to existing risk factors. Nonetheless, identifying underlying CVD in asymptomatic middle-aged individuals remains a key objective for primary care prevention as this offers an opportunity for education about lifestyle decisions and the potential for early treatment. Biomarkers offer the possibility of low-cost screening tools that may allow for better risk stratification and the identification of individuals who are not appropriately placed into risk categories using the current strategies of risk factor assessment.
In this regard, in the last years, a topic of research has been modeling to better predict CVD when these biomarkers are added to classical CVD risk factors such as blood pressure, age, smoking status, gender, BMI or lipids measures. Kaptoge, S. et al. showed that hs-CRP and fibrinogen improved the ability to predict CVD versus classical models [5]. McGranaghan p. et al. also found that adding metabolomic biomarkers improved risk prediction over the use of the classical risk factors only [55]. Additionally, in our study, three studies that included both hs-CRP and NT-proBNP-biomarkers for which metaanalyses were conducted-used new models with a combination of biomarkers, showing an improvement in risk assessment to predict CVD compared to when single biomarker or traditional risk assessment models were used [19,20,39].
The main limitation of the current work is that the data from the identified studies and pooled results were not sufficiently homogenous to make recommendations regarding the widespread use of serum biomarkers among middle-aged asymptomatic subjects to predict CVD in clinical use. Our results indicate an association between various biomarkers, specifically hs-CRP and NT-proBNP, and higher CV morbidity and mortality, which should guide further research, particularly with longer and comparable studies. In addition, some aspects of the search strategy could be viewed as a limitation of the current study, including only searching in one database (PubMed), using the English language exclusively, or specific terms because we could leave out other interesting biomarkers and important works. Methodological limitations include wide variability in study protocols among the studies reviewed and the failure of some of those studies to make appropriate adjustments in the statistical analysis for confounders such as statin use, body mass index, alcohol consumption or socioeconomic status. In addition, as our results showed, the length of follow-up of studies included in the meta-analysis varied, with the possibility of fewer events in shorter studies leading to an underestimation of the ability of these biomarkers to predict CVD compared with longer studies. On the other hand, the strength of the current work is that it has assessed the relation of potential biomarkers in a population that has not previously been well described but where disease prevention could potentially lessen cardiovascular diseases substantially. It is necessary to highlight the important gap in the role of biomarkers to predict CVD in this specific population despite better evidence for their use in secondary prevention and the elderly population.

Conclusions
Atherosclerosis starts at an early age as a low-grade systematic chronic inflammatory disease. The disease progression leads to clinical CVD with an elevated risk of an acute coronary event. The current work suggests the possibly important role of several biomarkers in improving the risk prediction of CVD in middle-aged asymptomatic individuals, especially for hs-CRP and NT-proBNP. However, longer and more extensive studies are needed to provide convincing evidence before general recommendations for clinical use can be made.