High-Sensitivity Cardiac Troponin T and Risk of Hypertension

Abstract

Background

Hypertension is often preceded by cardiac structural abnormalities. Thus, we assessed whether high-sensitivity cardiac troponin-T (hs-cTNT), a marker of chronic subclinical myocardial damage, can identify persons at risk for hypertension or left ventricular hypertrophy (LVH).

Methods and Results

We studied 6,516 ARIC Study participants, free of prevalent hypertension and cardiovascular disease at baseline (1990–1992). We examined the association of baseline hs-cTNT categories with incident diagnosed hypertension (defined by self-report of a diagnosis or medication use during a maximum of 19.9 years of follow-up) and with incident visit-based hypertension (defined by self-report, medication use, or measured BP >140/90 mmHg over 6 years). Relative to hs-cTNT <5ng/L, adjusted hazard ratios for incident diagnosed hypertension were 1.16 (95% CI 1.08, 1.25) for persons with hs-cTNT 5–8ng/L, 1.29 (1.14, 1.47) for hs-cTNT 9–13ng/L, and 1.31 (1.07, 1.61) for hs-cTNT ≥14ng/L (p-trend <0.001). Associations were stronger for incident visit-based hypertension. These associations were driven by higher relative hazard in normotensive persons (compared to those with prehypertension, p-interaction=0.001). Baseline hs-cTNT was also strongly associated with incident LVH by electrocardiography over 6 years (e.g. adjusted HR 5.19 [1.49–18.08] for hs-cTNT ≥14ng/L vs <5ng/L). Findings were not appreciably changed after accounting for competing deaths or adjustment for baseline BP levels or NT-proBNP.

Conclusions

In an ambulatory population with no history of cardiovascular disease, hs-cTNT was associated with incident hypertension and risk of LVH. Further research is needed to determine whether hs-cTNT can identify persons who may benefit from ambulatory BP monitoring or hypertension prevention lifestyle strategies.

Introduction

Hypertension remains a major cause of heart disease and stroke, with approximately 1 in 3 U.S. adults currently diagnosed as hypertensive and a further 6–10% of Americans estimated to have undiagnosed hypertension. ¹ In addition, a significant proportion of those with diagnosed hypertension have poorly controlled blood pressure (BP). ^{2, 3} Thus, there is a need to identify persons at risk either for progressing to clinically overt hypertension or for developing complications of this disease. In particular, BP control can be improved by more aggressive lifestyle and dietary changes in those with both prehypertension and overt hypertension. ⁴

The onset of hypertension is an insidious process that typically occurs over many years and is often preceded by altered diurnal BP patterns and prehypertension. ^{5, 6} While hypertension is diagnosed using defined BP thresholds, risks associated with elevated BP occur on a continuous spectrum. Indeed, even early, pre-hypertensive, BP abnormalities can induce structural heart changes⁷, potentially affording the opportunity both to detect persons at risk for overt hypertension and to initiate timely preventive strategies. ⁸ Prior studies have focused on the association between subclinical structural heart disease, particularly left ventricular hypertrophy (LVH), and subsequent development of hypertension. While these studies have confirmed that baseline LVH, measured by both echocardiography^{9, 10} and by cardiac MRI¹¹, is associated with incident hypertension, these imaging modalities are often impractical and too costly for routine screening.

New highly sensitive cardiac troponin T (hs-cTNT) assays have great potential as noninvasive laboratory-based markers of subclinical myocardial damage. ¹² These assays can detect subclinical myocardial damage in a significant proportion of persons who are free of known cardiovascular disease in the community. ^{13, 14} Interestingly, detectable levels of hs-cTNT appear to be more strongly linked to structural heart disease and its risk factors than to epicardial coronary artery disease. ^14–16 Thus, hs-cTNT screening may be of particular use in persons at risk for clinically apparent hypertension and other cardiovascular endpoints. Hs-cTNT may also help identify individuals who are at risk of developing hypertensive end-organ damage, such as LVH.

Thus, we sought to determine if hs-cTNT could identify persons at risk for subsequent hypertension or LVH in a large U.S. community-based cohort, the Atherosclerosis Risk in Communities (ARIC) Study.

METHODS

Study Population

The ARIC Study is a prospective cohort of 15,792 participants enrolled between 1987 and 1989 from four U.S. communities (Forsyth Country, North Carolina; Jackson, Mississippi; Minneapolis, Minnesota; and Washington County, Maryland). ¹⁷ We measured hs-cTNT in samples collected from all participants at ARIC visit 2, which took place from 1990 to 1992 and represents the baseline for this current analysis. Of the 14,348 persons who attended visit 2, we excluded all persons with prevalent diagnosed hypertension (n=6,801), those who had coronary heart disease (CHD, including silent myocardial infarction detected by electrocardiogram [ECG]), stroke, or heart failure at or prior to visit 2 (n=194), or those who were missing variables of interest (n=837) (eFigure 1). Thus, 6,516 persons were included in the main analysis evaluating our primary outcome of incident diagnosed hypertension (self-report of diagnosis or medication use during annual telephone follow-up). In our secondary analysis of incident visit-based hypertension, we excluded 599 persons with measured BP at baseline >140/90 mmHg (elevated blood pressure) and 242 persons without available BP measurement data, for an analytic study sample of 5,675 persons. In our analysis of incident LVH, we further excluded 44 persons with baseline LVH by ECG for an analytic sample of 5,631 (eTable 1). Institutional review boards at each clinical site reviewed the study and informed consent was obtained from all participants.

Measurement of hs-cTNT and other exposure variables

We measured hs-cTNT in stored serum samples, collected at visit 2, using the Roche high-sensitivity Troponin T reagent kit, a sandwich immunoassay performed on a Roche Elecsys 2010 Analyzer (Roche Diagnostics, Indianapolis, Indiana) at the University of Minnesota in 2012–2013. Intra-assay coefficients of variation (CVs) are 2.1% at a mean hs-cTNT concentration of 26 ng/L and 1.0% at 1990 ng/L. Inter-assay CVs are 6.0% at a mean hs-cTNT concentration of 25 ng/L and 3.7% at 1940 ng/L. The measurement range of the assay is 3– 100,000 ng/L, the limit of blank is 3 ng/L and the lower limit of detection is 5 ng/L. The intra-individual reliability of hs-cTNT values have been demonstrated to be high among ARIC study participants (laboratory reliability r value=0.94, six week biological reliability r value=0.94). ¹⁸ Participants self-reported smoking status. Plasma lipid concentrations and body weight and height for body mass index (BMI) were determined using standardized protocols. Glucose was measured using the hexokinase method.¹⁹ Diagnosed diabetes was defined as a self-reported physician diagnosis of diabetes or current use of diabetic medications. Glomerular filtration rate was estimated using serum creatinine and the CKD-EPI 2009 equation.²⁰ N-terminal pro-brain natriuretic peptide (NT-proBNP) was measured in stored serum samples from visit 2 on a Roche Elecsys 2010 Analyzer using a sandwich immunoassay method (Roche Diagnostics, Indianapolis, Indiana).

Follow-up for outcomes of interest

Participants were contacted annually via telephone, with follow-up currently available through March 2012. Incident diagnosed hypertension was assessed during these annual telephone calls using the following questions: “Has a doctor ever said you had high blood pressure?”, or “Since we last contacted you has a doctor said you had high blood pressure?”, and “Did you take any medications during the past two weeks for high blood pressure?”. After visit 2, ARIC participants also completed two further follow-up study examinations, visit 3 (1992–1995) and visit 4 (1996–1998). During these follow-up study visits, BP was recorded as the mean of at least 2 seated measurements using a manual random-zero sphygmomanometer.

We also evaluated the association between hs-cTNT and a well-known sequela of hypertension, LVH. This secondary outcome was assessed at visits 3 and 4, in persons free of both hypertension and LVH at visit 2, using resting 12-lead electrocardiograms and defined by Cornell criteria. ²¹

Statistical Analyses

Characteristics for the study population were presented according to categories of hs-cTNT (<5, 5–8, 9–13, or ≥14 ng/L) at baseline (visit 2, 1990–1992). While the hs-cTNT assay used can measure concentrations of troponin as low as 3 ng/L¹⁴, values <5 ng/L are measured with reduced precision and 5 ng/L represents the typical lower cut-point of detection in categorical analyses. Values ≥14ng/L represent approximately the 90th percentile of the ARIC population and correspond to the 99th percentile value for a “healthy” reference group of persons aged 20–70 years. ²² We further divided persons with values between 5–13 ng/L into two categories using a cutoff at the mid-point of this range. ¹⁵

The primary outcome of interest was incident diagnosed hypertension during a maximum of 19.9 and a median of 12 years of follow-up. Incident cases were identified at ARIC study visits 3 and 4 and, thereafter, by annual telephone contact with all participants. Cases were defined as a self-reported diagnosis of hypertension or antihypertensive medication use after visit 2. Time of incident diagnosed hypertension was the date when participants first reported diagnosis of or treatment for hypertension. We evaluated visit-based hypertension during a maximum of 8.7 and a median of 6 years follow-up as a secondary outcome. Visit-based cases of hypertension were identified by mean systolic BP ≥140mmHg or mean diastolic BP ≥90mmHg, self-reported physician diagnosis, or medication use at visits 3 (1993–1995) or 4 (1996–1998). For both hypertension endpoints, we also conducted analyses stratified by baseline BP category (normotensive [BP ≤120 mmHg systolic and ≤80 mmHg diastolic] or pre-hypertensive [BP >120 mmHg <140 mmHg systolic and/or >80 mmHg <90 mmHg]). Finally, we evaluated the association between baseline hs-cTNT and incident LVH at visit 3 (1993–1995) or visit 4 (1996–1998) among persons free from both hypertension and LVH at visit 2 (1990–1992).

We used Cox proportional hazards regression models for the primary outcome of diagnosed hypertension and discrete proportional hazards (cloglog) regression models for both secondary outcomes (visit-based incident hypertension or incident LVH) to estimate adjusted hazard ratios (HRs) for the association between baseline hs-cTNT and these outcomes. We modeled hs-cTNT as a categorical (<5 [reference], 5–8, 9–13, or ≥14 ng/L) and continuous exposure (log-transformed). In analyses with hs-cTNT as a continuous exposure, visit 2 hs-cTNT was modeled as untransformed or log-transformed and either truncated at <3ng/L (the limit of measurement) or with unmeasurable levels assigned a value of half the lower limit of measurement (1.5ng/L). ²³ Using Martingale residuals to assess fit, the log-transformed hs-cTNT without truncating fit the continuous models best. P-values for linear trend among hs-cTNT categories were obtained by assigning the median hs-cTNT value in the above categories of hs-cTNT and modeling this ordinal variable continuously. For all models, we verified the proportionality of the hazards with Schoenfeld residuals.

Models were adjusted for age (years), race-center (whites-Washington County; whites-Minneapolis; blacks-Jackson; blacks-Forsyth County, whites-Forsyth County), sex (male or female), body mass index (kg/m²), smoking (current; former; never), LDL-cholesterol (mg/dL), HDL-cholesterol (mg/dL), triglycerides (mg/dL), estimated glomerular filtration rate (mL/min/1.73m²), current lipid-lowering medication use (yes or no), LVH (yes or no), and diagnosed diabetes (yes or no). We tested for interactions by age, sex, and race. Model discrimination was assessed using Harrell’s C-statistic²⁴, and we evaluated improvement in the C-statistic for the addition of hs-cTNT as a log-transformed continuous variable to the fully adjusted model. Furthermore, for the primary outcome of diagnosed hypertension, we calculated the continuous net reclassification index (NRI) statistic ²⁵, also based on the addition of hs-cTNT to the full model.

In sensitivity analyses we additionally adjusted for baseline (visit 2) BP or NT-proBNP levels. Because baseline hs-cTNT is associated with all-cause mortality, cardiovascular death, and sudden cardiac death^{14, 16, 26–28}, we also conducted sensitivity analyses with a competing risk regression method (Fine-Gray approach) using a cumulative incidence function to account for intervening deaths. ²⁹

In addition, we also modeled hs-cTNT using linear splines in fully-adjusted Cox models for both incident diagnosed and incident visit-based hypertension, with knots at hs-cTNT concentrations of 5, 8 and 13 ng/L (figures truncated at the 99^th percentile), These models are shown graphically and overlaid on histograms showing the distribution of hs-cTNT in the study population. Finally, we repeated the analyses evaluating the outcome of incident diagnosed hypertension, after further excluding the 599 persons with elevated blood pressure at baseline. All analyses were performed using Stata version 13.0 (College Station, TX: StataCorp LP).

Results

In our community-based study sample of subjects without a history of cardiovascular disease and free from baseline hypertension, persons with higher baseline hs-cTNT were more likely to be older, male, black, pre-hypertensive, obese, diabetic, and have reduced kidney function (Table 1). Persons excluded due to missing values were not markedly different but were somewhat older, more likely to be black, and had a slightly more adverse health profile compared to those who were included in our study (eTable2).

Table 1.

Characteristics of the Study Population (ARIC participants without cardiovascular disease or diagnosed hypertension): Overall and According to categories of high sensitivity cardiac troponin T (ng/L) at baseline (1990–1992).

	Overall	Stratified by hs-cTNT
		<5 ng/L	5–8 ng/L	9–13 ng/L	≥14 ng/L
Number (%)	6516	4681 (72%)	1317 (20%)	384 (6%)	134 (2%)
Age, years	56 (6)	55 (5)	58 (6)	59 (6)	59 (6)
Male %	44	36	61	74	76
Black %	17	16	18	25	24
Current smoker %	23	25	18	17	20
Systolic blood pressure mmHg	115 (15)	114 (15)	118 (16)	120 (15)	124 (19)
Diastolic blood pressure mmHg	70 (9)	69 (9)	71 (9)	71 (9)	71 (10)
Hypertension categories (%)
Normotension   (BP <120/80 mmHg)	66	70	60	55	49
Prehypertension   (BP 120-139/80-89 mmHg)	27	25	32	35	35
Undiagnosed hypertension   (BP >140/90mmHg)*	6	5	8	10	16
LVH %	1	1	1	2	3
BMI, kg/m2
Normal weight % (< 25)	39	41	33	33	30
Overweight % (25 – 30)	41	40	45	43	43
Obese % (> 30)	20	19	22	25	28
Total cholesterol, mg/dL	207 (38)	208 (38)	206 (37)	203 (38)	206 (41)
LDL-cholesterol, mg/dL	132 (36)	132 (36)	133 (35)	131 (36)	131 (37)
HDL-cholesterol, mg/dL	52 (17)	53 (17)	48 (16)	47 (16)	49 (19)
Triglyceride, mg/dL	121 (61)	119 (59)	126 (64)	123 (67)	126 (68)
Lipid Medicines %	3	3	3	2	1
Diagnosed diabetes %	4	3	5	10	15
eGFR <60 mL/min/1.73m2 (%)	1	0	1	1	4

Estimates are mean (SD) or %, unless otherwise indicated. Hs-cTNT= high-sensitivity Troponin-T, BP= Blood Pressure, LVH= left ventricular hypertrophy, BMI= body mass index, LDL= Low Density Lipoprotein, HDL= High Density Lipoprotein, eGFR= estimated Glomerular Filtration Rate.

^*This group was excluded for the visit-based hypertension outcome analysis

During a maximum of 19.9 and a median 12 years of follow-up, 68% (n=4421) of the study sample developed our primary endpoint; incident diagnosed hypertension. Of these, 70% (n=3108) had hs-cTNT <5 ng/L at baseline, 21% (n=936) had hs-cTNT 5–8 ng/L, 6% (n=281) had hs-cTNT 9–13 ng/L, and 2% (n=96) had hs-cTNT ≥14 ng/L (Table 2). Crude incidence rates (per 1,000 person/years) of diagnosed hypertension were 54 in those with hs-cTNT <5 ng/L, 64 with hs-cTNT 5–8 ng/L, 73 with hs-cTNT 9–13 ng/L, and 74 in persons with baseline hs-cTNT ≥14 ng/L.

Table 2.

Crude incidence rates and adjusted^* hazard ratios (95% confidence intervals) for incident hypertension outcomes, according to baseline high sensitivity cardiac troponin T.

				Proportional Hazards Regression†	Competing Risks Regression‡
Baseline hs-cTnT	N	Events (n)	Incidence rate, per 1,000 person years (95% CI)	HR (95% CI)	HR (95% CI)
Incident Diagnosed Hypertension
Categories
<5 ng/L	4,681	3,108	54.0 (52.2–56.0)	1 (reference)	1 (reference)
5–8 ng/L	1,317	936	64.0 (60.0–68.3)	1.16 (1.08–1.25)	1.15 (1.06–1.24)
9–13 ng/L	384	281	72.6 (64.6–81.6)	1.29 (1.14–1.47)	1.21 (1.05–1.38)
≥14 ng/L	134	96	73.9 (60.5–90.2)	1.31 (1.07–1.61)	1.15 (0.91–1.44)
p-value for linear trend				<0.001	<0.001
Continuous
Log(hs-cTnT)	6,516	4421	57.2 (55.5–58.9)	1.14 (1.09–1.19)	1.11 (1.06–1.16)
Incident Visit-based Hypertension
Categories
<5 ng/L	4,139	1,059	47.1 (44.4–50.0)	1 (reference)	1 (reference)
5–8 ng/L	1,130	351	58.6 (52.8–65.1)	1.15 (1.01–1.30)	1.14 (1.00–1.29)
9–13 ng/L	306	120	74.9 (62.6–89.5)	1.32 (1.08–1.61)	1.38 (1.13–1.67)
≥14 ng/L	100	41	77.8 (57.3–105.6)	1.49 (1.09–2.05)	1.47 (1.07–2.01)
p-value for linear trend				<0.001	<0.001
Continuous
Log(hs-cTnT)	5,675	1,571	51.3 (48.9–53.9)	1.14 (1.06–1.22)	1.15 (1.07–1.24)

^*Adjusted for age (years), race-center (whites-Washington County; whites-Minneapolis; blacks-Jackson; blacks-Forsyth County, whites-Forsyth County), sex (male or female), body mass index (kg/m2), smoking (current; former; never), LDL-cholesterol (mg/dL), HDL-cholesterol (mg/dL), triglycerides (mg/dL), estimated glomerular filtration rate (mL/min/1.73m2), current lipid-lowering medication use (yes or no), left ventricular hypertrophy (yes or no), diagnosed diabetes (yes or no). Abbreviations as per Table 1.

^†Cox regression for diagnosed hypertension outcome, cloglog regression for visit-based hypertension

^‡Fine-Gray regression model. ²⁹ There were 532 interval deaths for the diagnosed hypertension outcome and 57 interval deaths for the visit-based hypertension outcome prior to administrative censoring.

After multivariable adjustment, baseline hs-cTNT remained significantly associated with incident diagnosed hypertension (Table 2). Specifically, relative to those with undetectable hs-cTNT at baseline, persons in the higher categories of hs-cTNT had a higher adjusted risk of hypertension. Similar results were found for log-hs-cTNT modeled as a continuous variable. A continuous and roughly linear association between baseline hs-cTNT and incident diagnosed hypertension is shown graphically in Figure 1a. Furthermore, for the primary outcome of incident diagnosed hypertension, the addition is hs-cTNT as a log-transformed continuous variable to our fully adjusted base model resulted in a significant increase in the C-statistic from 0.609 to 0.613 (p=0.003) and an improvement in continuous NRI of 20% (p=0.001). Results for the incident diagnosed hypertension outcome were not appreciably different in the sensitivity analysis that excluded persons with elevated BP at baseline (eTable 3).

Figure 1.

Adjusted* hazard ratio (95% confidence interval) of A) incident diagnosed hypertension and B) incident visit-based hypertension, according to baseline hs-cTnT modeled as linear splines (knots at 5, 8 and 13 ng/L) with background histogram of hs-cTNT in the study sample of ARIC subjects without baseline cardiovascular disease or hypertension. *Adjusted for age (years), race-center (whites-Washington County; whites-Minneapolis; blacks-Jackson; blacks-Forsyth County, whites-Forsyth County), sex (male or female), body mass index (kg/m2), smoking (current; former; never), LDL-cholesterol (mg/dL), HDL-cholesterol (mg/dL), triglycerides (mg/dL), estimated glomerular filtration rate (mL/min/1.73m2), current lipid-lowering medication use (yes or no), left ventricular hypertrophy (yes or no), diagnosed diabetes (yes or no). Abbreviations as per Table 1. Linear splines of hazard for hypertension with background distributional histogram of baseline hs-cTNT levels, truncated at the 99^th percentile of hs-cTnT (18 ng/L). Persons with hs-cTNT <3 ng/L were imputed values of 1.5 ng/L. Note that the "Proportion of the Population" axis label identifies the percentage of the ARIC analytic sample at each point on this background histogram. The shaded area around the regression line represents the 95% confidence interval.

The association between baseline hs-cTNT (both by category and as a continuous exposure) and our secondary outcome of incident visit-based hypertension was stronger than that for incident diagnosed hypertension (Table 2, Figure 1b). Further, in analyses stratified by baseline BP category (normotensive or pre-hypertensive), we observed that the overall association between hs-cTNT and both incident hypertension outcomes appeared to be largely driven by more robust associations in persons who were normotensive at baseline (p-for-interaction=0.04 for diagnosed hypertension and p-for-interaction=0.001 for visit-based hypertension) (Table 3 and eTable 4). Associations between hs-cTNT and both incident diagnosed and visit-based hypertension were largely unchanged after further adjusting for baseline BP or NT-proBNP (eTable 5). Interactions for the association of hs-cTNT and hypertension by age, gender, and race were all non-significant (all p-values-for-interaction >0.10).

Table 3.

Crude incidence rates and adjusted^* hazard ratios (95% confidence intervals) for incident diagnosed hypertension, according to baseline categories of high sensitivity cardiac troponin T: further stratified by baseline blood pressure (N=6,516).

				Proportional Hazards Regression†	Competing Risks Regression‡
Baseline hs-cTNT	N	Events (n)	Incidence rate, per 1,000 person years (95% CI)	HR (95% CI)	HR (95% CI)
NORMOTENSIVE SUBGROUP (BP ≤120 mmHg systolic and ≤80 mmHg diastolic)
Categorical
<5 ng/L	3,259	1,892	42.4 (40.6–44.4)	1 (reference)	1 (reference)
5–8 ng/L	785	480	47.9 (43.8–52.3)	1.13 (1.02–1.25)	1.12 (1.01–1.25)
9–13 ng/L	213	134	53.1 (44.8–62.9)	1.24 (1.03–1.48)	1.18 (0.97–1.43)
≥14 ng/L	65	41	54.1 (39.9–73.5)	1.39 (1.01–1.90)	1.29 (0.93–1.79)
p-value for linear trend				<0.001	0.006
Continuous
Log(hs-cTnT)	4,322	2,547	44.0 (42.3–45.7)	1.13 (1.06–1.20)	1.11 (1.04–1.18)
PREHYPERTENSION SUBGROUP (BP >120 <140 mmHg systolic and/or >80<90 mmHg diastolic)
Categorical
<5 ng/L	1,179	988	87.4 (82.1–93.0)	1 (reference)	1 (reference)
5–8 ng/L	424	356	90.6 (81.7–100.6)	1.11 (0.98–1.26)	1.11 (0.98–1.25)
9–13 ng/L	133	113	99.7 (82.9–119.9)	1.28 (1.05–1.58)	1.20 (0.97–1.48)
≥14 ng/L	47	33	78.6 (55.9–110.6)	0.88 (0.61–1.25)	0.72 (0.48–1.08)
p-value for linear trend				0.125	0.667
Continuous
Log(hs-cTnT)	1,783	1,490	88.8 (84.4–93.4)	1.07 (0.99–1.15)	1.03 (0.96–1.11)
ELEVATED BLOOD PRESSURE SUBGROUP (BP ≥140 mmHg systolic and/or ≥90mmHg diastolic)
Categorical
<5 ng/L	243	228	138.1 (121.3–157.3)	1 (reference)	1 (reference)
5–8 ng/L	108	100	150.1 (123.4–182.6)	1.14 (0.89–1.46)	1.10 (0.84–1.44)
9–13 ng/L	38	34	158.5 (113.2–221.8)	1.31 (0.90–1.93)	1.12 (0.76–1.65)
≥14 ng/L	22	22	180.1 (118.6–273.5)	1.45 (0.91–2.32)	1.53 (1.04–2.26)
p-value for linear trend				0.045	0.088
Continuous
Log(hs-cTnT)	411	384	144.7 (130.9–159.9)	1.14 (1.00–1.30)	1.10 (0.96–1.26)

^*Adjusted for age (years), race-center (whites-Washington County; whites-Minneapolis; blacks-Jackson; blacks-Forsyth County, whites-Forsyth County), sex (male or female), body mass index (kg/m2), smoking (current; former; never), LDL-cholesterol (mg/dL), HDL-cholesterol (mg/dL), triglycerides (mg/dL), estimated glomerular filtration rate (mL/min/1.73m2), current lipid-lowering medication use (yes or no), left ventricular hypertrophy (yes or no), diagnosed diabetes (yes or no). Abbreviations as per Table 1.

^†Cox regression for diagnosed hypertension outcome

^‡Fine-Gray regression model. ²⁹ There were 532 interval deaths for the diagnosed hypertension outcome prior to administrative censoring.

There were 532 competing deaths for the primary outcome of incident diagnosed hypertension over 12-years median follow-up, but only 57 competing deaths for the secondary outcome of incident visit-based hypertension over 6-years. In competing risk models, the association between hs-cTNT and incident diagnosed hypertension remained significant but was somewhat weakened in the hs-cTNT 5–8 ng/L and 9–13 ng/L categories and was no longer significant in those with baseline hs-cTNT ≥ 14 ng/L (Table 2). In contrast, results for the 6-year incident visit-based hypertension were similar in the Fine-Gray competing-risk models

Finally, we found that, relative to hs-cTNT of <5ng/L, persons in higher categories of hs-cTNT had a highly significant 6-year risk of incident LVH by ECG: HR, 2.29 (95% CI 1.24–4.26) for hs-cTNT 5–8ng/L, HR, 2.94 (95% CI 1.14–7.58) for hs-cTNT 9–13ng/L and HR, 5.19 (95% CI 1.49–18.08) for hs-cTNT ≥14ng/L (Table 4).

Table 4.

Adjusted^* Hazard Ratios (95% confidence intervals) for the secondary outcome of incident left ventricular hypertrophy over 6 years of follow-up by baseline high sensitivity cardiac troponin T among persons free of baseline cardiovascular disease or hypertension (N=5,631

			Proportional Hazards Regression†
Baseline hs-cTnT	N	Events (n)	HR (95% CI)
Categorical
<5 ng/L	4109	30	1 (reference)
5–8 ng/L	1121	17	2.29 (1.24–4.26)
9–13 ng/L	302	6	2.94 (1.14–7.58)
≥14 ng/L	99	3	5.19 (1.49–18.08)
p-value for trend			<0.001
Continuous
Log(hs-cTnT)	5,631	56	1.79 (1.27–2.52)

^*Adjusted for age (years), race-center (whites-Washington County; whites-Minneapolis; blacks-Jackson; blacks-Forsyth County, whites-Forsyth County), sex (male or female), body mass index (kg/m2), smoking (current; former; never), LDL-cholesterol (mg/dL), HDL-cholesterol (mg/dL), triglycerides (mg/dL), estimated glomerular filtration rate (mL/min/1.73m2), current lipid-lowering medication use (yes or no), diagnosed diabetes (yes or no). Abbreviations as per Table 1.

^†Cloglog regression for incident LVH by ECG

Discussion

In this large community-based study of U.S. adults, free of baseline hypertension and cardiovascular disease, we found that hs-cTNT was independently associated with subsequent development of hypertension, defined either by diagnosed cases (self-reported diagnosis or medication use) or by objective office measurement of BP, and also with incident LVH by ECG. Findings were similar by age, race and gender categories. Our results suggest that chronic subclinical myocardial damage, detected by elevated hs-cTNT, may precede the development of hypertension in the general population and that this novel biomarker of cardiac damage may have utility for identifying persons at future risk for hypertension and hypertensive end-organ damage.

Prior studies have reported the cross-sectional association between established hypertension and prevalent hs-cTNT elevations^{16, 30–32}. However, to our knowledge, this is the first study to evaluate the association between preceding subclinical myocardial damage, as measured by hs-cTNT, and subsequent risk of hypertension over prospective follow-up. This association was strongest for persons who had baseline normal BP readings (in contrast to those with pre-hypertension). Notably, the vast majority of individuals with prehypertension go on to develop hypertension, irrespective of baseline hs-cTNT. Therefore, the absolute risk of hypertension is very high in this entire sub-group, helping to explain why the relative impact of hs-cTNT for prediction of hypertension appeared non-significant in those with prehypertension.

While the exact pathway linking hs-cTNT to incident hypertension is unclear, a number of considerations are worth discussing. Elevated BP develops over time along a spectrum from normotension, to abnormal diurnal patterns of BP, to pre-hypertension, and then, ultimately, to clinical hypertension⁵. It is known that abnormalities in cardiac structure can occur long before the diagnosis of overt hypertension is made in the clinic^{5, 9, 33}. Based on our results, it is possible that these early BP aberrations cause myocardial damage, leading to elevated hs-cTNT, before clinically overt hypertension is recognized. Compatible with this hypothesis, it has been shown elsewhere that “non-dippers” (an abnormal diurnal BP pattern where sleep systolic BP fails to decrease >10% from daytime systolic BP and a known risk-factor for subsequent hypertension⁶), are more likely to have prevalent detectable troponin levels by hs-cTNT assays³⁴.

In addition, BP is highly variable and its true hemodynamic load cannot easily be captured by single office measurements (which are subject to misclassification error). Thus, it is also possible that the association between hs-cTNT and subsequent hypertension in ARIC could be a manifestation of the well-known phenomenon of “masked hypertension”, the clinical condition in which a patient’s office BP level is <140/90 mm Hg but home or ambulatory readings are in the hypertensive range³⁵. Masked hypertension occurs in over 10% of the adult population^{35, 36} and is known to result in end organ damage^{37, 38}.

Further, we found that elevated hs-cTNT was also a risk factor for the development of LVH in persons without baseline hypertension or clinical cardiovascular disease (including silent MI). This suggests that hs-cTNT may also predict future sequelae of hypertension and is consistent with our hypothesis that elevated hs-cTNT may reflect occult BP abnormalities in some individuals. Indeed, prior case-control data suggest that hs-cTNT is also associated with subsequent albuminuria in hypertensive patients. ³⁹ It is also worth noting that hs-cTNT could potentially be contributing more directly to the development of hypertension. For example, elevated hs-cTNT has been linked with arterial stiffness in diabetics⁴⁰.

Our results were robust to multiple sensitivity analyses. Standard Cox proportional-hazards models assume that persons censored (including those censored for death) prior to the end of follow-up have the same probability of incident hypertension as similar persons who remain under-observation. However, in persons who die, the development of hypertension cannot be observed. In this context, the Fine-Gray regression results, which account for interval deaths, are important to determine the ‘real-world’ impact of hs-cTNT screening in persons at risk for hypertension. In these Fine-Gray models, we found that the association between hs-cTNT and incident diagnosed hypertension (an endpoint with longer follow-up and increased interval accrual of deaths) was attenuated, particularly in the ≥14 ng/L category.

Our findings may have implications for future clinical practice. Specifically, normotensive persons with elevated hs-cTNT are at risk for subsequent hypertension and may benefit from evaluation for masked hypertension and from more intensive lifestyle BP interventions to reduce the likelihood of developing this morbid disease. Further research is necessary to confirm whether hs-cTNT could be useful in screening those at risk for hypertension.

This analysis has some limitations. This study was observational and, thus, may be subject to residual confounding. The small proportion of persons who were excluded for missing values (12%) had a slightly worse health profile than those included in our study. We do not have ambulatory BP recordings to determine whether or not masked hypertension or abnormal diurnal values were present in persons with elevated hs-cTNT. Cases of diagnosed hypertension were identified by self- report of diagnoses or medication used during annually follow-up; however, similar results were obtained in our analyses of visit-based hypertension (incorporating objectively elevated BP obtained during the clinical visit). Strengths of the study include the large sample size, bi-racial population, and rigorous measurement of cardiovascular risk factors.

In conclusion, elevated hs-cTNT in the general population, particularly among persons with normal blood pressure, identifies persons at risk for subsequent hypertension or LVH. Further studies are necessary to determine whether subclinical myocardial damage in these settings is due to masked hypertension or abnormal diurnal variability in BP. Elevated hs-cTNT in low-risk ambulatory populations may prove to be clinically useful in identifying persons at risk for hypertension, affording the opportunity both to consider ambulatory blood pressure monitoring and to initiate more intensive preventive strategies.