Low concentrations of cTnT, measured with a highly sensitive assay, were associated with a gradient of risk for new-onset HF and cardiovascular death in ambulatory community-dwelling individuals aged 65 years or older, independent of clinical variables associated with risk, as well as the cardiovascular risk biomarkers CRP and NT-proBNP. Furthermore, in this population, low cTnT concentrations are shown to frequently change over time. Independent of other risk factors as well as the baseline level of cTnT, these changes are associated with dynamic changes in risk of HF and cardiovascular death, concordant with the direction of change in biomarker level.
These findings expand upon prior studies using the highly sensitive cTnT assay outside the acute coronary syndrome setting in 3 ways. First, cTnT concentrations are detectable and of prognostic value in nearly two-thirds of a large geographically and ethnically diverse, stable, but at-risk population of ambulatory older individuals without a prior diagnosis of HF. The lower prevalence of detectable levels in the general population would be expected compared with studies of patients with stable coronary artery disease or symptomatic HF.14,15
Second, baseline levels of cTnT, below the range that would be expected to be detected with conventional assays, strongly associate with incident HF and cardiovascular death, independent of standard risk prediction variables. Addition of baseline cTnT to risk factor–adjusted models modestly improved discrimination, as measured by the C statistic and the IDI. Third, changes in cTnT during 2 to 3 years in older adults who remain free of HF, even when occurring at concentrations well below the 99th percentile of healthy younger blood donors,13
are prognostically significant.
These results should be considered in the context of prior findings with both conventional troponin T and I assays and other biomarkers for stratifying cardiovascular risk in community-dwelling older adults. In previous studies of older individuals, detectable levels of cardiac troponins were present in approximately 4% to 8%, and predictive of increased risk of HF, cardiovascular, and all-cause mortality.9–11
Similar to studies in patients with chronic but stable cardiovascular disease, the application of the highly sensitive cTnT assay increased the proportion of community-dwelling older adults with detectable cTnT levels approximately 10-fold.14,15
Compared with studies that used conventional troponin assays, the markedly increased range of measureable cTnT in our study enables estimation of a gradient of risk across the majority of older individuals, including those with an absence of clinical risk factors (other than age), and also permits examination of the significance of changing cTnT concentrations.
It is not possible from our study to determine the pathophysiology that results in detectable levels and frequent changes over time of cTnT in older adults. Prior evidence has shown that exercise-induced cardiac ischemia can lead to transient very low level increases in cardiac troponin levels as measured by a highly sensitive troponin I assay.29
Ischemia from known or unknown coronary artery disease must be considered in an older population, but magnetic resonance imaging in another stable older population does not support chronic ischemic heart disease as a predominant etiology linking low levels of troponins with subsequent development of HF.30
Furthermore, in a younger population with stable coronary artery disease small increases in cTnT levels measured by the highly sensitive cTnT assay were not predictive of myocardial infarction.15
Our findings also show that cTnT remains predictive of both HF and cardiovascular death in a subgroup of older adults with an absence of traditional risk factors, clinical history of heart disease, or an abnormal left ventricular ejection fraction (eTable 2).
Measurement of cTnT by the highly sensitive assay for risk stratification of older individuals has unique performance characteristics compared with other biomarkers that have been advocated by some for risk stratification in general populations, such as CRP and NT-proBNP. Although CRP has been associated with cardiovascular risk in younger populations, its prognostic value in older populations can be attenuated or absent.8,9,31
In contrast, natriuretic peptides, particularly NT-proBNP, perform better in higher-risk populations, including older adults compared with a younger general population.32
Our findings suggest that very low levels of cTnT provide prognostic information with respect to new-onset HF and cardiovascular mortality that is independent of NT-proBNP and CRP levels. The associations with HF were independent of other biomarkers such as renal function and electrocardiographic evidence of left ventricular hypertrophy.
The increment in the C statistic achieved by adding baseline cTnT to other clinical risk predictors is statistically significant, although more modest than in prior studies. In our analysis, we used as reference models outcome-specific clinical prediction models optimized for risk prediction in these individuals.24,25
The inclusion of variables such as albumin and left ventricular hypertrophy did optimize the clinical model but is beyond the traditional risk factors recommended for inclusion in statistical models when assessing the additional prognostic value of novel biomarkers.33
As a consequence, these reference models resulted in higher prognostic accuracy (C statistic, 0.75–0.78), even before adding cTnT than that reported in previous studies.9,15
Unique to our study is the finding that changes in very low levels of cTnT are common in this cohort of older adults and are independently associated with change in risk of both new-onset HF and cardiovascular death. Although it is possible that such changes reflect normal biological variation, the fact that these changes are associated with significant relative and absolute changes in risk regardless of baseline levels suggests that in fact they may represent a dynamic change in disease progression. Previously, we identified in the same population that change in NT-proBNP levels over time augmented prognostication above a single baseline measure.6
Taken together, these 2 findings suggest that serial measurements of both NT-proBNP and cTnT may improve risk assessment in elderly individuals.
Ultimately, the clinical importance of monitoring changes in cTnT levels for risk of progression to symptomatic HF is yet to be determined. This needs to be specifically considered in light of the conflicting findings with regard to improvement in discrimination, with significant improvement in one measure (the IDI) but not another (C statistic). However, the observation that changes in cTnT track with risk of HF and cardiovascular death reflects the dynamic nature of cardiovascular risk in older adults. Further studies are needed to assess whether monitoring low levels of cTnT may provide an opportunity to motivate specific changes in lifestyle or prompt medical interventions before progression to symptoms or cardiac structural abnormalities and to track the outcomes associated with these interventions.
Our study also has several limitations. First, samples were available in approximately three-fourths of the cohort at baseline, and differential absence of cTnT measures may have introduced bias into the estimates of associations with HF and cardiovascular death. Second, the duration of follow-up is a strength of our study; however, cardiovascular therapy has changed over time and it is possible that more ubiquitous use of medications such as statins could blunt the predictive value of the cTnT level. Third, unmeasured and residual confounding may have influenced our results; however, we demonstrate that cTnT concentration provides incremental prediction for HF and cardiovascular death beyond that provided by risk factors commonly used in clinical practice. Fourth, our choice of a more than 50% change in cTnT over time was based on biological variability in younger adults.26
Biological variability for cTnT using the highly sensitive cTnT assay has been reported to be higher in a small cohort of younger adults, for whom cTnT concentrations were mostly less than 3.00 pg/mL (the limit of detection of the current version of the highly sensitive cTnT assay).34
Therefore, these results may not be as relevant to our analysis. Biological variability in older adults with greater comorbidities remains to be determined.