The primary prevention of cardiovascular (CV) events has achieved a great deal, but there is much room for further improvement. In practice, primary prevention focuses on achieving target levels of longstanding risk factors such as blood pressure (BP) and cholesterol and then awaiting new symptoms due to overt heart disease. The inadequacy of this current approach can be illustrated in 3 ways. First, CV events are still the most common cause of hospitalizations and death. Second, it has been estimated that 43% of coronary events would still occur even if we achieved perfect risk factor control (1). Third, 40% to 50% of sudden cardiac deaths (SCDs) occur in individuals before they have ever developed overt heart disease (2). In fact, during the 50- to 59-year decade, 1% of all men suffer an SCD before ever developing overt heart disease (2).

Those who die suddenly with no overt heart disease do, however, often have silent cardiac target organ damage (cTOD). For example, patients with silent ischemia are known to have a 21-fold increase in risk of a coronary event (3), and coronary artery disease (CAD) is the most common autopsy finding in cases of SCD (4). However, we not only want to reduce SCDs; we also want to better prevent nonfatal CV events, and silent ischemia is not the only predictor of nonfatal events. Tsang et al. (5) have also shown that echocardiographic abnormalities like left ventricular hypertrophy (LVH), left ventricular systolic dysfunction (LVSD), left ventricular diastolic dysfunction (LVDD), and left atrial enlargement (LAE) each independently predict CV events (hazard ratios of 1.42 to 1.70] (5).

Yet the key weakness in current primary prevention is that it is not standard practice to investigate primary prevention patients to identify silent cTOD, because “phenotyping” all primary prevention patients would be prohibitively expensive. What might reduce the costs of phenotyping would be if treated primary prevention patients could be pre-screened with a biomarker that identifies those patients most likely to be harboring silent cTOD and thereafter to target only them for detailed cardiac phenotyping. There are several potential biomarkers for this—B-type natriuretic peptide (BNP), high-sensitivity cardiac troponin T (hs-cTnT), microalbuminuria, estimated glomerular filtration rate (eGFR), uric acid, or some combination of these. Therefore, we set out to see how well each biomarker would identify cTOD in a population of treated primary prevention patients. If positive, this could validate a possible way forward to improve primary prevention, by using a screening biomarker to select patients for cardiac phenotyping.

This was a cross-sectional study. The subjects in this study were randomly recruited between April 2008 and July 2010 from local general practitioner surgeries and from the CV risk clinic at Ninewells Hospital, Dundee, United Kingdom. In sampling from both sources, random suitable patients were written to so that the populations recruited were unbiased from within each source. From the list of those who responded to the initial invitation, recruitment was from consecutive patients. The response rate was 52%, and there were no major differences in the characteristics of the 2 cohorts or the overall results. Patients were suitable if they were 50 years of age or above and eligible for primary prevention only with no previous known CV disease. They had to be stable on therapy for at least ≥1 year and to have reached target for their primary risk factor (e.g., office BP ≤140/90 mm Hg or 25% reduction in total cholesterol) (9). We excluded those with previously known CV disease, known renal impairment (eGFR <60 ml/min), atrial fibrillation, and significant (defined as more than mild) valvular heart disease either on auscultation or echocardiography. All study subjects underwent clinical assessment, biochemical measurements (including BNP, hs-cTnT, and urinary microalbumin), electrocardiography (ECG), transthoracic echocardiography, dobutamine stress echocardiography (DSE) to detect myocardial ischemia, and 24-h ambulatory BP measurement (Spacelab Healthcare, Issaquah, Washington).

Biochemical measurements including BNP and hs-cTnT were made by trained laboratory staff blinded to clinical and echocardiographic data. The BNP was measured with Triage BNP assay (Biosite, Inc., San Diego, California), whereas hs-cTnT was measured with a highly sensitive assay on an automated platform (Elecsys E170, Roche Diagnostics, Basel, Switzerland) with a lower limit of detection of 3 ng/l. Microalbuminuria was defined as urinary albumin/creatinine ratio of >30 mg/g, measured on a spot urine sample, whereas eGFR was calculated with Modification of Diet in Renal Disease formula.

A comprehensive echocardiogram was performed in each individual including M-mode, 2-dimensional, color, and tissue Doppler imaging. All echocardiograms were performed by a single trained operator (M.A.N.) with the same echocardiographic instrument (Philips iE33, Philips Healthcare, Andover, Massachusetts) according to a standardized protocol based on the recommendations by the American Society of Echocardiography (10). The operator was blinded to the biochemical data, including BNP and hs-cTnT. LVH was defined as left ventricular (LV) mass index >95 g/m² in women and LV mass index >115 g/m² in men as per the agreed American Society of Echocardiography criteria (10). Left atrial volume (LAV) was calculated with ellipsoid formula, and an LAV index of >28 ml/m² was used to define LAE. Left ventricular systolic dysfunction was defined as (modified Simpson's rule) EF <50%, and LVDD was defined as mitral (lateral annulus) E/e' of >15. Those with a borderline E/e' (9) were considered to have LVDD if they also had an enlarged left atrium as suggested (11). Interobserver and intraobserver agreements for echocardiographic measures were performed by re-analyzing echocardiographic images of randomly selected study subjects. With alpha model reliability analysis, we calculated intra-class correlation coefficient with a 95% confidence interval (CI). The intraclass correlation coefficient for M-mode, 2-dimensional, and Doppler imaging ranged between 0.93 and 0.98 for the intraobserver variability and between 0.92 and 0.96 for the inter-observer variability. The presence of myocardial ischemia was primarily assessed by DSE according to a standard protocol (12). A 16-segment model was used and new or worsening regional wall motion abnormalities of ≥1 LV segment were considered as indicative of inducible ischemia. Those patients where DSE was inconclusive for technical reasons underwent tetrofosmin myocardial perfusion scanning. Dipyridamole was used as the stressor with gated single-photon emission computed tomography analysis.

Continuous variables are reported as median (interquartile range [IQR]), and categorical variables are reported as proportions. Characteristics of patients with or without cTOD were compared (Table 1) by the chi-square test for categorical variables and by the Student t test or Mann-Whitney U test for continuous variables as appropriate. The biomarkers were analyzed as continuous as well as ordered categorical variables. Receiver-operating characteristic (ROC) curves were constructed for the biomarkers to assess their ability to identify presence or absence of cTOD as a primary outcome measure. The area under the ROC curves was compared by a nonparametric approach. The incremental value of combining various biomarkers was calculated by multivariate logistic regression. We also divided the study cohort into tertiles on the basis of BNP levels according to a pre-specified protocol. This tertile approach has previously been taken by the Framingham Heart Study group (13), and we pre-specified this tertile-based analysis at the time of study conception. The significance level for the trend across the tertiles was calculated by Jonckheere-Terpstra test and chi-square test. A similar analysis was performed for hs-cTnT. All statistical analyses were performed with SPSS for Windows (version 16.0, SPSS, Chicago, Illinois), and a 2-sided p value of <0.05 was considered to be significant. The net reclassification index was calculated as described by Pencina et al. (14). The Tayside Research and Ethics Committee approved the research protocol, and all study participants provided written informed consent. Power calculations at the outset by our statistician suggested we needed 83/tertile to detect an 8% LVH prevalence in the low BNP tertile versus a 30% prevalence in the high tertile. We rounded these figures up to 100/tertile. We assessed the 2 cohorts (hospital and general practitioner) separately, but both results were essentially the same.

A total of 300 patients were included in this study (Table 1). They had been receiving therapy for their primary risk factor for a mean duration of 3.3 ± 2.8 years (40.7% treated hypertensives, 49% both treated hypertension and dyslipidemia, and 10.3% treated dyslipidemia only). Ultrasound contrast agent was used in 11% of patients, and myocardial perfusion scans were used in 7%. A total of 102 (34%) participants had evidence of at least 1 form of cTOD. Left ventricular hypertrophy was the most prevalent (29.7%) form of cTOD, followed by LVDD (21.3%), LAE (15.3%), LVSD (6.3%), and ischemia (6.3%). Of those with cTOD, 29% had 1 form of cTOD, 31% had 2 forms, 29% had 3, and 10% had 4 or more forms of cTOD. Only 1 patient had evidence of a previous unrecognized myocardial infarction and no patients had any significant right ventricular abnormalities.

The BNP levels were significantly higher (median: 25.4 [IQR: 15.6 to 42.4] pg/ml vs. 10.6 [IQR: 6.3 to 18.8] pg/ml, p < 0.001) in those with cTOD compared with those without (Figure 49_gr1A). In a multivariate logistic model that included age, sex, body mass index, and eGFR, BNP (p < 0.001) remained an independent predictor of underlying cTOD. A 1-SD rise in log BNP was associated with an increased risk of existing cTOD (adjusted odds ratio: 3.3, 95% CI: 2.3 to 4.7). Moreover, BNP levels were higher in those who had multiple (≥2) forms of cTOD compared with those who had a single form of cTOD (median: 33.2 [IQR: 19.2 to 56.1] pg/ml vs. 15.8 [IQR: 10.8 to 23.5] pg/ml, p < 0.001) (Figure 49_gr2). The AUC for BNP to identify any form of cTOD was 0.78 (95% CI: 0.73 to 0.83, p < 0.001). We re-calculated the AUC after BNP was adjusted for age, sex, eGFR, and body mass index, but no significant change in the AUC (p = 0.198) was noted. We also recalculated the AUC for BNP by stratifying study population by age groups. The AUC was better (0.82 vs. 0.77) among those younger than 60 years compared with those ≥60 years old. The BNP also seemed to perform better in men (Table 2). When divided into tertiles, the prevalence of each cTOD significantly (p < 0.001) increased from the bottom BNP tertile to the top BNP tertile (Figure 49_gr1A).

Similarly, hs-cTnT levels were significantly higher (median: 5.91 [IQR: 3.97 to 8.95] ng/l vs. 3.72 [IQR: 3.00 to 5.93] ng/l, p < 0.001) in those with cTOD compared with those without. In multivariate logistic analysis, after correcting for age, sex, and eGFR, hs-cTnT (p < 0.001) remained an independent predictor of existing underlying cTOD. A 1-SD rise in logarithmically transformed hs-cTnT was associated with an increased risk of existing cTOD (adjusted odds ratio: 2.1, 95% CI: 1.55 to 2.82). The AUC for hs-cTnT to identify any form of cTOD was 0.70 (95% CI: 0.63 to 0.76, p < 0.001). When divided into tertiles, the prevalence of each cTOD (except LVSD) increased significantly from the bottom hs-cTnT tertile to the top hs-cTnT tertile (Figure 49_gr1B).

Microalbuminuria (defined as urinary albumin/creatinine ratio of >30 mg/g) was an independent predictor of cTOD, but its AUC was only 0.61. Similarly, both uric acid and eGFR had poor discriminating power, with AUC 0.49 (p = 0.43) and 0.58 (p = 0.02), respectively (Table 3).

We classified ECG as normal, minor abnormalities, and major abnormalities as described by Davie et al. (15). In our study, neither ECG parameter performed well (Table 3). Clinical prediction scores (Framingham, QRISK, and ASSIGN) also performed poorly (AUCs of 0.60, 0.51, and 0.62), but their performance was greatly enhanced by adding BNP (Table 4). The net reclassification index was 11% for BNP alone and 12.3% for BNP plus hs-cTnT in comparison with Framingham. When BNP and hs-cTnT were combined, the AUC was 0.81, and the number of missed cases was low (Table 5).

We have found that BNP screening of treated primary prevention patients is an effective way to identify which treated primary prevention patients are already harboring silent cTOD. Its test-performing characteristics are similar to other commonly used screening tests, including prostate-specific antigen (PSA) for prostatic cancer (AUC 0.68 to 0.78), mammography for breast cancer (AUC 0.78), and pap smears for cervical cancer (AUC 0.74) (16). The performance of hs-cTnT was not as good as BNP and the other biomarkers, but the combination of BNP and hs-cTnT was best.

Previous work in various different populations had already shown that BNP could identify 1 or 2 isolated abnormalities like LVH, LAE, and LV dysfunction and silent ischemia (19). However, it is worth emphasizing the novelty and importance of this study. Firstly, no previous BNP study had looked so comprehensively at all of these forms of cTOD in the 1 study. This is very important, because patients harboring forms of cTOD not assessed in prior studies will appear as false positives for BNP unless the search for cTOD is comprehensive, as here. Indeed, it is likely to be the comprehensive nature of our cTOD phenotyping in this study that has led to our AUCs being so good and in particular to our key observation that BNP is at least doubled when more than 1 form of cTOD exists. Secondly, all diagnostic and screening tests perform differently in different populations, and no previous study had examined (treated) primary prevention patients. Thirdly, we were able to directly compare BNP and hs-cTnT in a single population, and no previous publication has compared both biomarkers in the 1 population. However, the main and unique importance of this work is that we have now validated a possible new way forward (BNP ± cTnT screening and selected phenotyping) to improve the primary prevention of CV events.

In our study, BNP performed better in men, as seen before (24). Also, hs-cTnT levels were closely associated with LV mass and LA volume, which is in line with previously published data from a general population (25).

It is intriguing that in the published data, novel plasma biomarkers (like BNP) on their own have only modestly improved ROC curves for CV risk prediction over traditional risk factors (26). Our suggested approach is BNP followed by targeted phenotyping, and this is very different from using BNP on its own, because the patients who would be regarded as being at high risk by our suggested approach would have had to fail both a biomarker test and specific investigations for cTOD. Furthermore, another major benefit of our suggested approach is that the intensified treatment in each patient could be targeted toward the exact form of cTOD found in that individual (i.e., personalized medicine) and not just to some generally increased risk identified by a biomarker, which by itself does not indicate the source of the high risk.

This does beg the question of what additional therapies the identification of cTOD might lead to. For silent ischemia, additional therapies could be beta blockade, aspirin, and statins: neither aspirin nor beta-blockade are routinely recommended for primary prevention per se (beta blockers have been relegated to 4th line antihypertensives), and statins are only used in a proportion. One study did suggest that such an anti-ischemic regime could reduce events by 80% in silent ischemia (27). It could even lead to coronary angioplasty in selected cases, where 2 studies in silent ischemia have produced impressive results (28). As to the finding of normotensive LVH, new treatments could be extra BP reduction, allopurinol or copper chelation, or even the preferential use of ARBs or aldosterone antagonists, because they have a better evidence base that they are effective at regressing LVH (30). As to the finding of LVSD, the addition of beta blockers, angiotensin-converting enzyme inhibitors, and aldosterone antagonists in combination could markedly reduce risk, possibly by 40% to 50% according to the trials in early heart failure. We deliberately focused on the cardiac abnormalities that are known to be harmful rather than just those known to be currently treatable, because future treatments are bound to be developed. For example, the TOPCAT (Trial of Aldosterone Antagonist Therapy in Adults With Preserved Ejection Fraction Congestive Heart Failure) study might endorse spironolactone for LVDD. Left atrial dilation predicts atrial fibrillation, and in the future we might be able to use aldosterone blockade to prevent atrial fibrillation developing, as occurred in the EMPHASIS-HF (Eplerenone in Mild Patients Hospitalization and Survival Study in Heart Failure) study.

It is obviously important to consider the possible cost-effectiveness of BNP screening followed by targeted phenotyping. It is worth emphasizing that the phenotyping only involves a baseline echo and a stress echo that could be done in 1 session. Furthermore, all the drug interventions proposed in the preceding text are cheap, nongeneric drugs. In this analysis, we have to focus on the top BNP tertile only group, because it is only for this group that we have detailed information on outcomes, from Paget et al. (35). A rough calculation for 300 patients might be £3,000 ($4,500 USD) for 300 BNPs, £2,0000 ($3,000 USD) for approximately 100 echocardiograms using published costs, and £70/year ($105/year USD) for aspirin, and so forth—in the patients who fail both the BNP test and the echocardiogram test (36). In terms of effect, we would expect 8 deaths over 7 years in our top tertile (35). If we assume our treatment reduces those by 25% (and the SWISS I study [Swiss Interventional Study on Silent Ischemia type I] suggests it could be up to 80% for some patients), then we could expect to save 2 lives in 7 years. Overall this would mean that each life saved costs approximately £25,000 ($37,500 USD) over 7 years (i.e., £3,500/year [$5,250/year USD]). If extended to 11 years, the cost would be less for each life saved (i.e., £19,111 [$28,666 USD] each or £1,737/year [$2,605/year USD]). Of course this analysis only addresses total deaths and not the nonfatal CV events that our strategy is likely to also reduce. Fewer nonfatal CV events could also lead ultimately to fewer hospital stays for heart failure and chest pain, which will also enhance the cost-effectiveness of this approach.

Even without further targeting to specific patient subgroups, it seems that BNP overall is a little better at screening than PSA. At a level of 1.1 ng/ml, PSA is 83% sensitive and only 39% specific, whereas at 3.1 ng/ml, it is 32% sensitive and 86% specific. If these PSA figures are compared with BNP cutoffs of 10 and 30 pg/ml, BNP outperforms PSA as a screening test (Table 2). The PSA is probably the best comparator for BNP, because both are continuous variables. The reason PSA has not been adopted is not because PSA is a poor screening test but rather because prostate cancer can behave benignly, whereas its treatments are fairly invasive.

We would not be rigid, from these data, on which exact group should be phenotyped (e.g., the top BNP tertile group [>20 pg/ml] only or the group with a BNP >15 pg/ml or a cTnT >5.93 ng/l or some other group). Further larger studies should address which exact cohort should be phenotyped. Another key issue is the number of cTOD cases missed by any cutoff, and it is clear from (Table 5) that we would only miss 13% of cTOD cases if we used 1 of the best-performing cutoffs (BNP >15 pg/ml or cTnT >5.93 ng/l). However, even this figure overstates its significance, because the cases missed are seldom the more serious cases, (i.e., those with LVSD or silent ischemia or ≥2 forms of cTOD are very seldom missed [0% to 7%]) (Figure 49_gr2).

The main limitation of this study is that some forms of cTOD were fairly uncommon in our population. The main example is silent ischemia (6.3%), although the latter still showed a clear increment in frequency across the BNP tertiles (i.e., 1% in bottom and middle tertiles to 17% in top BNP tertile, which agrees overall with earlier data) (20). Intriguingly, the equivalent figures for hs-cTnT were less discriminating at 2%, 8%, and 9% across the tertiles. (Table 3) shows, however, that our results are very similar even if we exclude silent ischemia. Another limitation is that we do not yet have outcome data on our study population, but a recent study in a very similar primary prevention population found a 3-fold increase in mortality at the top versus the bottom tertile of N-terminal pro-BNP, even after adjusting for traditional risk factors (35). Indeed, the accompanying editorial said that we now need to know in this population what increases BNP in the absence of ECG-LVH (37). Our study answers that very question.

Cardiovascular disease and cancer are our main causes of death, and both often have a pre-symptomatic stage. As a result, screening programs have been developed to identify early target organ disease in cancer, but such screening programs for target organ disease do not yet exist in cardiology, even though sudden death can occur in the pre-symptomatic stage of heart disease. However, BNP ± cTnT screening now seems to be able to identify those who already have silent cTOD in a treated population of primary prevention patients with similar accuracy to other commonly used screening tests, such as those for cancer. Furthermore, the cases missed are seldom the more serious forms of cTOD (i.e., those with LVSD or ischemia or ≥2 cTODs). If our results are confirmed, this could lead to trials of adding BNP (± hs-cTnT) screening plus targeted phenotyping to primary prevention to see whether it really can deliver better primary prevention of CV events in a cost-effective way. If so, this could one day propel screening for pre-symptomatic CV disease into the same league as screening for certain cancers achieved long ago.