Contribution of 30 Biomarkers to 10-Year Cardiovascular Risk Estimation in 2 Population Cohorts
The MONICA, Risk, Genetics, Archiving, and Monograph (MORGAM) Biomarker Project
Background— Cardiovascular risk estimation by novel biomarkers needs assessment in disease-free population cohorts, followed up for incident cardiovascular events, assaying the serum and plasma archived at baseline. We report results from 2 cohorts in such a continuing study.
Methods and Results— Thirty novel biomarkers from different pathophysiological pathways were evaluated in 7915 men and women of the FINRISK97 population cohort with 538 incident cardiovascular events at 10 years (fatal or nonfatal coronary or stroke events), from which a biomarker score was developed and then validated in the 2551 men of the Belfast Prospective Epidemiological Study of Myocardial Infarction (PRIME) cohort (260 events). No single biomarker consistently improved risk estimation in FINRISK97 men and FINRISK97 women and the Belfast PRIME Men cohort after allowing for confounding factors; however, the strongest associations (with hazard ratio per SD in FINRISK97 men) were found for N-terminal pro-brain natriuretic peptide (1.23), C-reactive protein (1.23), B-type natriuretic peptide (1.19), and sensitive troponin I (1.18). A biomarker score was developed from the FINRISK97 cohort with the use of regression coefficients and lasso methods, with selection of troponin I, C-reactive protein, and N-terminal pro-brain natriuretic peptide. Adding this score to a conventional risk factor model in the Belfast PRIME Men cohort validated it by improved c-statistics (P=0.004) and integrated discrimination (P<0.0001) and led to significant reclassification of individuals into risk categories (P=0.0008).
Conclusions— The addition of a biomarker score including N-terminal pro-brain natriuretic peptide, C-reactive protein, and sensitive troponin I to a conventional risk model improved 10-year risk estimation for cardiovascular events in 2 middle-aged European populations. Further validation is needed in other populations and age groups.
Received August 11, 2009; accepted March 15, 2010.
Cardiovascular risk assessment based on classic risk factors does not fully explain the distribution of risk in the general population.1–4 Specifically, classic risk scores may provide variable results in different populations.5 Recent data indicate that 9 simple risk factors, including abnormal apolipoprotein levels, smoking, diabetes mellitus, and hypertension, substantially account for the risk of acute myocardial infarction globally.6 Importantly, most of these classic risk factors are modifiable, and intervention is likely to reduce the risk of cardiovascular disease (CVD). To improve risk estimation beyond what is possible with classic risk factors, many novel biomarkers have now been related to cardiovascular risk in community settings.7 It seems that, overall, C-reactive protein and N-terminal pro-brain natriuretic peptide (NT-proBNP) have most consistently improved risk estimates in specific populations.8–10 Nevertheless, it remains unclear whether inclusion of any individual biomarker in a model based on classic risk factors results in more accurate risk assessment across different populations and whether any modification in risk estimation warrants a change in treatment.
Editorial see p 2381
Clinical Perspective on p 2397
Simultaneous assessment of multiple biomarkers may add clinically useful information.11,12 As yet, no study has explored the possibility of refining cardiovascular risk estimation with the use of an extensive set of novel biomarkers assessed in large prospective studies in different countries. Thus far, on the basis of 2 large-scale studies of initially healthy individuals, the Reynolds Risk Score was developed, indicating that the addition of C-reactive protein and parental history significantly improves global cardiovascular risk prediction.13,14 To address the issue of multiple biomarkers, we evaluated the incremental value of 30 biomarkers reflecting most known pathways for CVD compared with more readily available simple risk factors for cardiovascular risk estimation in 10 466 individuals, with 798 incident CVD events, enrolled in 2 European cohorts. We derived a simple biomarker score in the FINRISK97 study and validated its power for cardiovascular risk assessment in the Northern Irish Belfast Prospective Epidemiological Study of Myocardial Infarction (PRIME) Men cohort.
MORGAM Study Population
The multinational MONICA, Risk, Genetics, Archiving, and Monograph (MORGAM) (http://www.ktl.fi/morgam) project was initially established to develop cardiovascular risk scores based on classic risk factors and to determine whether genetic variability or biomarker assessment could improve on them. An overall summary of the project’s objectives and detailed descriptions of the cohorts have been published elsewhere.15,16 The MORGAM Biomarker Study includes all MORGAM Participating Centers with blood samples for biomarker assessment. The main end point is the first occurrence of a major cardiovascular event during follow-up, which includes the first fatal or nonfatal definite or possible myocardial infarction or coronary death, unstable angina, cardiac revascularization, ischemic stroke, and unclassifiable death. Details of the event classification are provided in Table I in the online-only Data Supplement and in Reference 16.
Cohort Description of FINRISK97 and Belfast PRIME Men
The design of the FINRISK97 study is described elsewhere.17 Briefly, the population sample, examined in 1997, included people aged 25 to 74 years. The survey included a self-administered questionnaire and measurements of height, weight, and blood pressure. A venous blood specimen was taken to determine serum total cholesterol, triglycerides, and high-density lipoprotein (HDL) cholesterol. Serum and plasma samples were stored for the analysis of biomarkers at −70°C. Follow-up to the end of 2007 was performed with the use of record linkage of the FINRISK data with 3 data sources: (1) FINAMI myocardial infarction register,18 (2) National Hospital Discharge Register, and (3) National Causes of Death Register. The latter 2 are countrywide, computerized registers that cover all hospitalizations and deaths in Finland. The cardiovascular diagnoses in these registers have been validated recently.19
Belfast PRIME Men Cohort
The design of the PRIME Men trial is described elsewhere.20 Briefly, PRIME Men is a population-based prospective study of coronary events. During the 1991–1994 recruitment period, men aged 50 to 59 years were examined in Belfast. The entry examination included standardized questionnaires, measurement of weight, height, and blood pressure, and blood sampling for lipid measurements. Serum and plasma samples for the analysis of biomarkers were stored at −80°C.
During the 10-year follow-up, subjects were contacted annually by letter and, if necessary, by telephone, and a clinical event questionnaire was completed. For all possible events, clinical information was sought directly from hospital or general practitioners’ notes. All details of ECGs, hospital admissions, enzymes, surgical operations, angioplasties, and treatment were collected, and death certificates were obtained for supporting information on cause of death. The events were classified according to standard criteria.21 After 10 years, follow-up was achieved in >98% of the cohort.
For the present analyses, we excluded participants with a history of major CVD such as myocardial infarction, hospitalized unstable angina, coronary artery bypass grafting, percutaneous transluminal coronary angioplasty, or ischemic or hemorrhagic stroke. In FINRISK, we checked the National Hospital Discharge Register for earlier events and used either documented or self-reported history of a major cardiovascular event as an exclusion criterion. In Belfast, self-reported history was used. Details of the FINRISK97 and Belfast PRIME Men cohorts, as well as an assessment of the quality of the data, have been published elsewhere.16,21 In both studies, approval from the local research ethics committee was obtained, and all subjects gave written informed consent.
We selected 30 biomarkers representing lipid metabolism, inflammation, hemodynamic physiology, vascular function, oxidative stress, coagulation, renal function, angiogenesis, and myocardial necrosis (Figure 1). All biomarkers were measured in the MORGAM Mainz Biomarker Laboratory. For detailed analytical description, including the assay type, the material used (EDTA plasma or serum), and intra-assay and interassay coefficients of variation, see Table II in the online-only Data Supplement. Here, we present the details of the 4 analytes of interest for deriving a biomarker score: C-reactive protein was determined by latex immunoassay CRP16 (Abbott, Architect c8000), with intra-assay and interassay coefficients of variation of 0.93 and 0.83. NT-proBNP was determined by the electrochemiluminescence sandwich immunoassay ECLIA (Roche Diagnostics, Elecsys 2010), with intra-assay and interassay coefficients of variation of 2.58 and 1.38. BNP was determined by the chemiluminescent microparticle immunoassay CMIA (Abbott, Architect i2000), with intra-assay and interassay coefficients of variation of 2.11 and 4.28. For troponin I (STAT Troponin I immunoassay, Abbott Diagnostics), the detection range is 0 to 50 ng/mL, and the 10% coefficient of variation is 0.032 ng/mL, which is below the 99th percentile in the FINRISK population at 0.035 ng/mL and thus fulfills the criteria of a contemporary sensitive troponin assay. The intra-assay and interassay coefficients of variation were 2.36 and 4.80. All coefficients of variation had been obtained in the Mainz MORGAM Biomarker Laboratory. Additionally, serum lipids were measured in Helsinki and Lille, with external quality control by the World Health Organization Regional Lipid Reference Center in Prague.22 All biochemical analyses were performed with the investigators blinded to patient status.
The analysis comprised 3 stages. The first was an assessment of the associations between the single biomarkers and study end points; the second was an assessment of the discriminatory ability of single biomarkers in risk prediction models; and the third involved the derivation of a multiple biomarker score and its validation. In the first 2 stages, the assessment was undertaken primarily in the male cohort of FINRISK, and it was replicated in the FINRISK women and the men of the Belfast PRIME cohort. In the third stage, the derivation of the score was based on the FINRISK male cohort, and it was validated in the Belfast PRIME Men cohort.
In all statistical analyses, missing biomarker data were handled with the use of multiple imputation.23 Details of missing biomarker measurements are provided in Table III in the online-only Data Supplement. Outcome information was included in the imputation model to avoid attenuation of estimated effects in later analyses.24 Imputation was performed with the use of WinBUGS software. Some biomarkers included measurements above or below the detection limits, and, in these cases, we used the detection limits in place of the undetectable values. In all time-to-event models, we used cubic root transformation for biomarker distributions with skewness >2. The transformed variables are indicated in Table IV in the online-only Data Supplement. For multivariate model selection, we applied lasso penalized regression, with the penalty parameter selected by 10-fold cross-validation.25 Time-to-event models were fitted with the use of the “survival” package of R statistical software, and lasso model selection was performed with the use of the “penalized” package of R.26 The aim in model building was the prediction of 10-year absolute risk. When prediction models derived from the same data were compared, effects of overfitting were avoided by calculating the predicted risks with the use of 10-fold cross-validation. The discriminative ability of the models was tested with the use of c-index improvement, integrated discrimination improvement (IDI) statistics, and net reclassification improvement (NRI),27 the latter according to recently suggested risk limits indicating the relevant strata of 0% to <5%, 5% to <10%, 10% to <20%, and ≥20% for 10-year risk.28,29 Model calibration was tested with the Hosmer-Lemeshow test with 10 risk groups. To assess the sensitivity of the results on the used imputation model, analysis with the use of only subjects with complete information on biomarkers was also performed. This did not change the conclusions. For a more detailed statistical description, see Methods in the online-only Data Supplement.
Baseline Characteristics and Classic Risk Factors
After the exclusion of individuals with prevalent CVD, 3870 men and 4045 women were available for analysis from the FINRISK97 cohort and 2551 men from the Belfast PRIME Men cohort. Baseline characteristics and the median concentration of all biomarkers determined are shown in Table 1⇓. Levels of several markers within the respective biological system were highly correlated (Table V in the online-only Data Supplement). During the follow-up, 376 men and 162 women experienced an incident fatal or nonfatal cardiovascular event in the FINRISK97 cohort and 260 in the Belfast PRIME Men cohort.
Association of Single Biomarkers With Incident Cardiovascular Events
To assess the independent association between each biomarker and future cardiovascular events, we built a series of Cox regression models (Figure 2 and Table IV in the online-only Data Supplement). A number of variables were independently associated with future cardiovascular events in FINRISK97 men, with P<10−3 after we accounted for area, age, and classic risk factors including body mass index, smoking, systolic blood pressure, history of diabetes mellitus, non-HDL cholesterol, HDL cholesterol, and cardiovascular drugs. The hazard ratios (HRs) per 1-SD increment were as follows: NT-proBNP, HR=1.23 (95% confidence interval, 1.14 to 1.34; P=7.0×10−7); C-reactive protein, HR=1.23 (1.13 to 1.35; P=5.8×10−6); troponin I, HR=1.18 (1.09 to 1.28; P=7.9×10−5); D-dimer, HR=1.22 (1.10 to 1.34; P=9.5× 10−5); BNP, HR=1.19 (1.09 to 1.31; P=2.3×10−4). Associations for NT-proBNP, C-reactive protein, and troponin I were replicated in the PRIME Men cohort, and associations for NT-proBNP, C-reactive protein, and BNP were replicated in FINRISK97 women (Figure 2).
Single Biomarkers for Cardiovascular Risk Estimation
We performed a series of analyses to explore the incremental effect of single biomarker variables when added to the prediction models including the classic risk factors.
We computed the c-index associated with the risk-estimation model based on classic risk factors alone and compared it with the models based on a combination of the classic risk factors and each biomarker separately. Because of the wide age range, the basic model for the FINRISK97 male cohort already achieved high discrimination, with a c-index value of 0.817. Of those biomarkers strongly associated with future cardiovascular risk, only the addition of troponin I (0.821; P=0.009) and D-dimer (0.821; P=0.014) individually increased the c-index. However, these results were not replicated in the PRIME Men cohort. Overall, none of the single markers improved the c-statistics in all 3 cohorts simultaneously. By contrast, when the IDI was considered, only NT-proBNP achieved an improvement in model discrimination across all 3 cohorts. Detailed results of c-statistics and IDI are shown in Table 2 for the clinically relevant markers, and results of all markers are presented in Table VI in the online-only Data Supplement.
We assessed whether adding a biomarker to the model assigned an individual to a different risk category. Because of the wide age range of the FINRISK97 cohort, we attempted reclassification only for the Belfast PRIME Men cohort. The results are shown in Table VII in the online-only Data Supplement. Here, only NT-proBNP, C-reactive protein, leptin, and the apolipoprotein B100/apolipoprotein A-1 ratio improved net reclassification, with P<0.05.
Biomarker Score and Improvement in Cardiovascular Risk Estimation
Development of the Score
To assess whether the addition of multiple biomarkers to a basic model added any information for cardiovascular risk assessment, we derived a biomarker score using the data on FINRISK97 men. By Cox regression models on single biomarkers, the 3 strongest associations with biomarkers were for NT-proBNP (P=7.0×10−7), C-reactive protein (P=5.8×10−6), and troponin I (P=7.9×10−5). The same selection of the best biomarkers was obtained with the use of the lasso technique in a multivariate Cox regression model including all markers: On a 1-SD increment scale, HRs were 1.13 for NT-proBNP, 1.12 for C-reactive protein, and 1.10 for troponin I. The independent effect of these markers on risk was due to the fact that they were only weakly correlated with one another and reflect different aspects of cardiovascular pathology. Because no additional marker improved discrimination in FINRISK97 men when added to a prediction model that included NT-proBNP, C-reactive protein, and troponin I, we selected a score consisting of these 3 markers. The biomarkers in the score were weighted by their respective regression coefficients in a Cox regression model that also included the classic risk factors, fitted in the FINRISK97 men. For details of score development, see Table VIII in the online-only Data Supplement.
Validation of the Score
Application of the biomarker score in the Belfast PRIME Men validation cohort improved risk estimation. First, the score, considered a continuous variable, was strongly associated with the risk of incident CVD events (HR=1.36; 95% confidence interval, 1.22 to 1.52; P=1.4×10−8) (Figure 2). Second, the probability of event-free survival differed significantly between thirds of the biomarker score (Figure 3A). Third, in prediction models that included age, classic risk factors, and treatment variables, the ability of the model to discriminate events from nonevents was significantly improved after inclusion of the biomarker score (Figure 3B). Fourth, the addition of the biomarker score assigned a significant proportion of the population to a different risk category (Figure 3C). The NRI was estimated at 0.11 (P=0.0008), resulting from net 6% of cases classified up and net 5% of noncases classified down. If the 0% to 6% and 6% to 20% risk stratification classification was applied, the NRI was 7.3% (P=0.006). Similarly, the improvements in IDI and c-index (from 0.67 to 0.70) were highly significant (P<0.0001 and P=0.004, respectively). Finally, model calibration analyses did not show any significant deviation between predicted and observed risk (Figure 3D). Overall, these validation analyses suggest that the addition of the biomarker score that included C-reactive protein, NT-proBNP, and sensitive troponin I improved risk estimation and discrimination between incident cases and noncases.
Cut Point Analyses
To test applicable thresholds directly, we first explored data-derived optimal cutoff values in the FINRISK men on the basis of the IDI criterion. These were as follows: for CRP, 6.81 mg/L; for sensitive troponin I, 0.008 ng/mL; and for NT-proBNP, 187 pg/mL. These optimal cut points derived from the data were applied directly to the PRIME Men population. Table 3 describes HRs for the 3 single markers as well as the 3-marker score described above. We also experimented with a cutoff-based score, derived from FINRISK97 men and tested in the Belfast PRIME Men cohort. The prediction results with the threshold-based score were virtually identical to the results obtained for the score based on continuous variables: The c-index improved from 0.67 to 0.69 with the score (P=0.006), NRI was 11% (P<0.0001), and IDI was highly significant (P<0.0001). The NRI of the 0% to 6% and 6% to 20% risk stratification was somewhat lower (7.3%; P=0.006).
We measured 30 novel cardiovascular biomarkers in 2 prospective European population-based studies including 10 466 subjects and found that several of these biomarkers were independently associated with incident cardiovascular events over 10 years of follow-up. Only NT-proBNP, C-reactive protein, and sensitive troponin I were consistently associated with cardiovascular events across studies in both men and women. However, adding any single biomarker separately to the established risk model did not improve risk estimation in all 3 cohorts. By contrast, incorporation of a composite biomarker score including the noninterdependent variables troponin I, C-reactive protein, and (NT-pro)BNP significantly improved risk assessment. This was assessed by model discrimination and calibration and significantly improved the classification of the subjects into risk categories in the validation cohort.
Our data agree with other population-based studies that have shown an association between various biomarkers and incident cardiovascular events but have failed to demonstrate improved cardiovascular risk prediction on the basis of single biomarker assessment.30–32 Interestingly, a recent article by Melander and coworkers33 evaluated 6 novel biomarkers and found, consistent with our study, that C-reactive protein and N-terminal fragment of BNP were the best predictors of cardiovascular events. However, this study did not include troponin I and did not achieve a significant net reclassification improvement.
Each of the biomarkers included in our score reflects an important pathophysiological pathway in CVD, such as inflammation, disturbed hemodynamic function, and cardiac micronecrosis. These 3 analytes can be measured by robust, well-standardized, reproducible immunoassays that are available at reasonable cost. In particular, the reliable detection of very low levels of troponin by the use of sensitive, robust tests now makes testing feasible in community settings. Importantly, these biomarkers are unrelated, and each is therefore contributing independently and additively to risk assessment. We selected the biomarkers using regression analyses to identify those with the strongest association with cardiovascular end points in the 2 separate cohorts. We also used lasso analyses to achieve both parsimony and a reduction in estimation variance, and both methods resulted in identical biomarker selection. Most importantly, however, we derived the biomarker score in a large prospective, community-based cohort and validated it in a separate cohort. In doing so, we demonstrated significantly improved measures of discrimination, calibration, and reclassification. If data-derived cut points had been used for the 3 single markers C-reactive protein, NT-proBNP, and sensitive troponin I as well as for establishing the score, prediction results of the score would have been virtually identical to those obtained from the continuous variables. This indicates robustness of model specification.
Why are our findings at variance with those reported from other community-based studies?32 An obvious answer relates to the different biomarker combination. Inclusion of troponin I in the score substantially improves predictive ability. The importance of these measures has been postulated earlier.34 However, the present investigation for the first time applies a contemporary sensitive troponin with a 10% coefficient of variation at the 99th population percentile. Further efforts are needed to apply highly sensitive tests in population-based studies. Furthermore, the validation cohort has a narrow age range of 50 to 59 years, and age often has the single largest effect among the discrimination criteria. In other cohorts, younger individuals or individuals with a wider age range have been examined. Another recent investigation suggested that in elderly men, with and without CVD, the simultaneous addition of markers of renal function and CVD improved risk estimation for total mortality.12 However, these results were based on fatal events only, and the study combined elderly men with and without CVD. In contrast, the present data provide knowledge about the manner in which a robust biomarker combination can add information for cardiovascular risk stratification in population-based studies of initially healthy individuals. Based on the initial set of 30 biomarkers, our modified risk score has been derived in a large prospective, population-based study and validated in another cohort from a different European region that demonstrates some measure of generalizability.
Some limitations merit consideration. We measured all biomarkers from frozen samples, which may affect absolute risk estimates. We cannot address the question of whether a reduction of the biomarker score would result in reduction of the risk. Clearly, other novel biomarkers that we did not test might subsequently be found to improve the risk score. Thresholds of biomarkers and their variation over time need to be validated. In particular, the future use of more sensitive troponin measures will require adaptation and validation of the threshold. Furthermore, we cannot ascertain whether preventive treatment decisions based on the new score would improve the outcome.
In conclusion, the addition of NT-proBNP, C-reactive protein, and troponin I to an established risk model improves 10-year risk estimation for nonfatal and fatal cardiovascular events in middle-aged European populations. Whether this will lead to improved treatment decision making needs to be tested prospectively.
Sites and Key Personnel of the Contributing MORGAM Centers
FINRISK—National Institute for Health and Welfare, Helsinki: V. Salomaa (Principal Investigator), A. Juolevi, E. Vartiainen, P. Jousilahti. MORGAM Data Center—National Institute for Health and Welfare, Helsinki: K. Kuulasmaa (head), Z. Cepaitis, A. Haukijärvi, B. Joseph, J. Karvanen, S. Kulathinal, M. Niemelä, O. Saarela.
MORGAM Biomarker Laboratory—Johannes Gutenberg University, Mainz: S. Blankenberg, Tanja Zeller.
Belfast PRIME Men—Queen’s University Belfast, Belfast, Northern Ireland: F. Kee (Principal Investigator), A. Evans (former Principal Investigator), J. Yarnell, E. Gardner. MORGAM Coordinating Center—Queen’s University Belfast, Belfast, Northern Ireland: A. Evans (MORGAM coordinator), S. Cashman.
MORGAM Management Group
A. Evans (chair), S. Blankenberg, F. Cambien (Paris, France), M. Ferrario (Varese, Italy), K. Kuulasmaa, L. Peltonen (Helsinki, Finland), M. Perola (Helsinki, Finland), A. Peters (Munich, Germany), V. Salomaa, D. Shields (Dublin, Ireland), H. Tunstall-Pedoe, P.G. Wiklund (Umeå, Sweden).
We are indebted to Margot Neuser for graphic work.
Sources of Funding
The MORGAM Study in general was funded in part through the European Community’s Seventh Framework Program (FP7/2007-2013), ENGAGE project, grant agreement HEALTH-F4-2007-201413. The MORGAM Biomarker Study specifically is funded by the Medical Research Council London (G0601463, identification No. 80983: Biomarkers in the MORGAM Populations). Dr Salomaa was supported by the Finnish Foundation for Cardiovascular Research and by the Finnish Academy (grant 129494).
Abbott Diagnostics, BRAHMS AG, Diadexus, and Roche Diagnostics provided test reagents (for details, see Table I in the online-only Data Supplement). Dr Blankenberg has received honoraria from Abbott Diagnostics (<5,000 USD) and Roche Diagnostics (<5,000 USD) and has worked as a consultant/advisory board member for Brahms (<5,000 USD). The authors report no other conflicts.
Tolonen H, Mahonen M, Asplund K, Rastenyte D, Kuulasmaa K, Vanuzzo D, Tuomilehto J. Do trends in population levels of blood pressure and other cardiovascular risk factors explain trends in stroke event rates? Comparisons of 15 populations in 9 countries within the WHO MONICA Stroke Project: World Health Organization Monitoring of Trends and Determinants in Cardiovascular Disease. Stroke. 2002; 33: 2367–2375.
Conroy RM, Pyorala K, Fitzgerald AP, Sans S, Menotti A, De Backer G, De Bacquer D, Ducimetiere P, Jousilahti P, Keil U, Njolstad I, Oganov RG, Thomsen T, Tunstall-Pedoe H, Tverdal A, Wedel H, Whincup P, Wilhelmsen L, Graham IM. Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project. Eur Heart J. 2003; 24: 987–1003.
Vasan RS. Biomarkers of cardiovascular disease: molecular basis and practical considerations. Circulation. 2006; 113: 2335–2362.
Blankenberg S, McQueen MJ, Smieja M, Pogue J, Balion C, Lonn E, Rupprecht HJ, Bickel C, Tiret L, Cambien F, Gerstein H, Munzel T, Yusuf S. Comparative impact of multiple biomarkers and N-terminal pro-brain natriuretic peptide in the context of conventional risk factors for the prediction of recurrent cardiovascular events in the Heart Outcomes Prevention Evaluation (HOPE) Study. Circulation. 2006; 114: 201–208.
Ridker PM, Paynter NP, Rifai N, Gaziano JM, Cook NR. C-reactive protein and parental history improve global cardiovascular risk prediction: the Reynolds Risk Score for men. Circulation. 2008; 118: 2243–2251,4p following 2251.
Evans A, Salomaa V, Kulathinal S, Asplund K, Cambien F, Ferrario M, Perola M, Peltonen L, Shields D, Tunstall-Pedoe H, Kuulasmaa K. MORGAM (an international pooling of cardiovascular cohorts). Int J Epidemiol. 2005; 34: 21–27.
Kulathinal S, Niemelä M, Kuulasmaa K, contributors from Participating Centres, for the MORGAM Project. Description of MORGAM cohorts (2005). Available at: http://www.ktl.fi/publications/morgam/cohorts/index.html, URN:NBN:fi-fe20051214. Accessed May 5, 2010.
Vartiainen E, Laatikainen T, Peltonen M, Juolevi A, Mannisto S, Sundvall J, Jousilahti P, Salomaa V, Valsta L, Puska P. Thirty-five-year trends in cardiovascular risk factors in Finland. Int J Epidemiol. 2010; 39: 504–518.
Salomaa V, Ketonen M, Koukkunen H, Immonen-Raiha P, Lehtonen A, Torppa J, Kuulasmaa K, Kesaniemi YA, Pyorala K. The effect of correcting for troponins on trends in coronary heart disease events in Finland during 1993–2002: the FINAMI study. Eur Heart J. 2006; 27: 2394–2399.
Pajunen P, Koukkunen H, Ketonen M, Jerkkola T, Immonen-Raiha P, Karja-Koskenkari P, Mahonen M, Niemela M, Kuulasmaa K, Palomaki P, Mustonen J, Lehtonen A, Arstila M, Vuorenmaa T, Lehto S, Miettinen H, Torppa J, Tuomilehto J, Kesaniemi YA, Pyorala K, Salomaa V. The validity of the Finnish Hospital Discharge Register and Causes of Death Register data on coronary heart disease. Eur J Cardiovasc Prev Rehabil. 2005; 12: 132–137.
Yarnell JW. The PRIME study: classical risk factors do not explain the severalfold differences in risk of coronary heart disease between France and Northern Ireland: Prospective Epidemiological Study of Myocardial Infarction. QJM. 1998; 91: 667–676.
Niemelä M, Kulathinal S, Kuulasmaa K, eds, for the MORGAM Project. Description and quality assessment of MORGAM data (2007). Available at: http://www.ktl.fi/publications/morgam/qa/contents.htm, URN:NBN:fi-fe20071495. Accessed May 5, 2010.
Tunstall-Pedoe H, ed. Prepared by Tunstall-Pedoe H, Kuulasmaa K, Tolonen H, Davidson M, Mendis S, with 64 other contributors for the WHO MONICA Project. MONICA Monograph and Multimedia Sourcebook. Geneva, Switzerland: World Health Organization; 2003.
Rubin DB. Multiple Imputation for Nonresponse in Surveys. New York, NY: J Wiley & Sons; 1987.
R Development Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2006. Available at: http://www.R-project.org. Accessed May 5, 2010.
Cook NR. Use and misuse of the receiver operating characteristic curve in risk prediction. Circulation. 2007; 115: 928–935.
Melander O, Newton-Cheh C, Almgren P, Hedblad B, Berglund G, Engstrom G, Persson M, Smith JG, Magnusson M, Christensson A, Struck J, Morgenthaler NG, Bergmann A, Pencina MJ, Wang TJ. Novel and conventional biomarkers for prediction of incident cardiovascular events in the community. JAMA. 2009; 302: 49–57.
Zethelius B, Johnston N, Venge P. Troponin I as a predictor of coronary heart disease and mortality in 70-year-old men: a community-based cohort study. Circulation. 2006; 113: 1071–1078.
We examined 30 novel biomarkers representing different pathophysiological pathways in 7915 men and women of the population-based FINRISK97 cohort with 538 incident cardiovascular events at 10 years of follow-up. We then developed a biomarker score from the best biomarkers and validated the score in the 2551 men of the Belfast Prospective Epidemiological Study of Myocardial Infarction (PRIME) cohort with 260 incident cardiovascular events at 10 years of follow-up. The strongest associations with the risk for future cardiovascular events in FINRISK97 were found for N-terminal pro-brain natriuretic peptide, C-reactive protein, brain natriuretic peptide, and sensitive troponin I. In addition to pathophysiological considerations, application of different statistical methods consistently resulted in selection of troponin I, C-reactive protein, and N-terminal pro-brain natriuretic peptide for the establishment of the biomarker score. Adding this score to the conventional risk factor model in Belfast PRIME Men led to additional prognostic information beyond that obtained from the classic risk factors. Furthermore, inclusion of the biomarker score into established risk models significantly reclassified cardiovascular risk estimates by 11% (P=0.0008). This large prospective study in individuals without a history of major cardiovascular disease at baseline showed that the addition of a score consisting of 3 biomarkers to the conventional risk factor model improved the estimation of 10-year risk for serious cardiovascular events. Further validation is needed in other populations, ethnicities, and age groups.
Continuing medical education (CME) credit is available for this article. Go to http://cme.ahajournals.org to take the quiz.
↵*Drs Evans and Salomaa contributed equally to this work.
Sites and key personnel of the contributing MORGAM Centers are given in the Appendix.
The online-only Data Supplement is available with this article at http://circ.ahajournals.org/cgi/content/full/CIRCULATIONAHA.109.901413/DC1.