Association of Thromboxane Generation with Long-term Survival in Aspirin Users and Non-users

Abstract

BACKGROUND:

Persistent systemic thromboxane generation, predominantly from non-platelet sources, in aspirin (ASA) users with cardiovascular disease (CVD) is a mortality risk factor.

OBJECTIVES:

To determine the mortality risk associated with systemic thromboxane generation in an unselected population irrespective of ASA use.

METHODS:

Stable thromboxane metabolites (TXB₂-M) were measured by ELISA in banked urine from 3044 participants (mean age 66±9 years, 53.8% women) in the Framingham Heart Study. The association of TXB₂-M to survival over a median observation period of 11.9 years (IQR 10.6, 12.7) was determined by multivariable modeling.

RESULTS:

In 1363 participants (44.8%) taking ASA at the index examination, median TXB₂-M was lower than in ASA non-users (1147 vs. 4179 pg/mg creatinine, P<0.0001). TXB₂-M was significantly associated with all-cause and cardiovascular mortality irrespective of ASA use (HR 1.96 and 2.41, respectively, P<0.0001 for both) for TXB₂-M in the highest quartile based on ASA use compared to lower quartiles, and remained significant after adjustment for mortality risk factors for similarly-aged individuals (HR 1.49 and 1.82, respectively, P≤0.005 for both). In 2353 participants without CVD, TXB₂-M was associated with cardiovascular mortality in ASA non-users (adjusted HR 3.04, 95% CI, 1.29, 7.16) but not in ASA users, while ASA use was associated with all-cause mortality in those with low (adjusted HR 1.46, 95% CI 1.14, 1.87) but not elevated TXB₂-M.

CONCLUSION:

Systemic thromboxane generation is an independent risk factor for all-cause and cardiovascular mortality irrespective of ASA use and its measurement may be useful for risk stratification, particularly in those without CVD.

CONDENSED ABSTRACT

Persistent thromboxane generation in aspirin users with cardiovascular disease, originating predominantly from non-platelet sources, is associated with adverse outcome. In 3044 participants in the Framingham Heart Study, thromboxane generation was associated with long-term adjusted all-cause and cardiovascular mortality, irrespective of aspirin use. In participants without cardiovascular disease, thromboxane generation was associated with increased mortality in aspirin non-users whereas aspirin use was associated with increased mortality in those with low levels of thromboxane generation. These findings suggests that measurement of systemic TXA₂ generation may be useful in risk stratification, particularly in those without cardiovascular disease.

INTRODUCTION

Thromboxane A₂ (TXA₂) is an eicosanoid with potent platelet activating and vasoconstrictor properties generated from arachidonic acid by the actions of cyclooxygenase (COX) and downstream thromboxane synthase (TXAS) enzymes.¹ In healthy individuals, TXA₂ generation is thought to occur mainly in platelets and is effectively inhibited by aspirin (ASA), which irreversibly inhibits the COX-1 enzyme. In over a quarter of individuals with cardiovascular disease (CVD), especially those with increased oxidative stress, substantial amounts of TXA₂ is generated in non-platelet tissues that is not fully suppressed by standard ASA therapy.^2–5 Several modestly-sized studies of individuals with CVD taking ASA revealed that systemic TXA₂ generation was associated with adverse cardiovascular outcome and death, though the mechanism by which this occurs is not fully understood.^6–10 It is unknown if systemic TXA₂ generation from combined platelet and non-platelet sources in non-ASA users also predicts adverse clinical outcome or if the stimuli for systemic TXA₂ generation differ between ASA users and non-users.

To address these knowledge gaps, we measured systemic non-renal TXA₂ generation by quantifying stable metabolites of thromboxane B₂ (TXB₂-M)¹¹ in banked urine samples from participants in the Offspring and Omni cohorts of the Framingham Heart Study and determined its relationship to long-term survival. To help identify differential stimuli for systemic TXA₂ generation, we analyzed variables associated with systemic TXA₂ generation in participants stratified by ASA use at the time of the index examinations.

METHODS

Study Cohort:

The Framingham Heart Study (FHS) is a longitudinal community-based study comprised of several study cohorts with serial examinations every 4 to 8 years.¹² The study population for this investigation included participants in examination cycles 8 (2005–2008) of the Offspring and cycle 3 (2007 to 2008) of the Omni cohorts in whom there was an available urine sample banked at the time of the examination.^13,14 At these examinations, attendees underwent a medical history, a cardiovascular-focused physical examination, anthropometric and blood pressure measurements, and laboratory assessment of standard cardiovascular disease risk factors. Written informed consent was obtained by study participants at each examination and study protocols were approved by the human subject institutional review boards of the Boston University School of Medicine and the University of Massachusetts Chan Medical School.

Biologic Assays:

TXB₂-M was measured in duplicate in urine samples stored at −80ºC using the AspirinWorks^® 11dhTXB₂ Test Kit (Corgenix, Bromfield, CO) and normalized to urine creatinine according to the manufacturer’s instruction. The assay has a limit of detection of 156 pg/mL with a linear range up to 5000 pg/mL. Samples with values below the limit of detection were reported as 156 pg/mL while those above 5000 pg/mL were re-assayed at an additional 1:4 dilution and reported as 20,000 pg/mL if they remained above the linear range. The intra-assay coefficient of variance was 3.3%. The stability of TXB₂-M measurements with this assay in urine after prolonged storage and multiple freeze-thaw cycles was verified in our laboratory.¹⁵ Urine 8-iso-PGF_2α , serum C-reactive protein (CRP), interleukin-6 (IL-6), monocyte chemotactic protein (MCP), plasma P-selectin and lipoprotein-associated phospholipase A₂ (Lp-PLA₂) and other standard laboratory variables had previously been measured in these participants as described^16,17 with biomarker measurement manuals available on the FHS website at http://www.framinghamheartstudy.org.

Statistical Analyses:

Descriptive statistics were calculated as usual with ASA use defined by review of medications at the time of the examination. Frequency of categorical variables was compared between groups using the Pearson chi-square test. Means of continuous normally distributed variables were compared using Student’s t-test, while non-normally distributed variables were compared using the Kruskal Wallis test. Multivariable modeling was used to determine variables associated with urinary TXB₂-M after natural log transformation using a standard general linear model (GLM). Univariate modeling of variables by ASA use was first conducted to determine the significance of the effect on lnTXB₂-M. An additional variable was added to adjust for the ASA dose (dichotomized as a weekly average of ≤ 81 versus >81 mg/day). Variables with significant effects (p ≤ 0.05) on lnTXB₂-M were then included in a full GLM and backward elimination of non-significant variables (p-value >0.05) was used to achieve the final parsimonious model.

Cox proportional hazards regression was used to model the relationship between urinary TXB₂-M and time to death after confirming the assumption of proportionality was satisfied and accounting for an interaction between ASA use and urinary TXB₂-M where applicable. Urinary TXB₂-M was considered as a binary variable of respective quartile 4 versus quartiles 1–3 for ASA users and non-users based on the Kaplan-Meier survival plots. The Fine-Gray method was used to account for competing risk when analyzing causes of death.¹⁸ Multivariable analysis was used to adjust for relevant predictors of mortality as indicated.

RESULTS

Baseline Characteristics

Of the 3021 Offspring and 298 Omni 1 Cohort participants, urine samples were available and TXB₂-M measurable in 3044 (91.7%) participants, of whom 1363 (44.8%) were taking ASA at the time of the index examination. Median urinary TXB₂-M was expectedly lower in ASA users compared to non-users with some participants in both groups exhibiting very high levels (Figure 1). Characteristics of participants stratified by ASA use and in those without TXB₂-M determination are shown in Table 1. Compared to ASA non-users, ASA users were older, with a higher BMI, more likely to be male and have a higher prevalence of risk factors for/established CVD with associated medical therapy.

Figure 1. Thromboxane generation in aspirin users and non-users.

Systemic thromboxane generation was quantified by measuring urinary thromboxane B₂ metabolites (TXB₂-M) in 1681 aspirin (ASA) users and 1681 ASA non-users. Median values with interquartile rages (boxes), 5% and 95% confidence intervals (bars) and individual outliers are shown.

Table 1.

Clinical and laboratory characteristics of participants at the index examination stratified by aspirin use.

Characteristic	ASA Users (N = 1363)	ASA Non-users (N = 1681)	P-value	Missing TXB2-M (N=275)
Age, mean (SD)	68 (8)	64 (9)	<0.0001	73 (10)
Female sex, n (%)	604 (44.3)	1033 (61.5)	<0.0001	212 (77.1)
Non-white race, n (%)	84 (6.2)	124 (7.5)	0.1635	3 (1.1)
Hispanic ethnicity, n (%)	34 (2.8)	58 (3.9)	0.1416	9 (3.9)
BMI (kg/m2), mean (SD)	28.8 (5.4)	28.0 (5.5)	<0.0001	28.4 (5.6)
eGFR (mL/min/1.73 m2), mean (SD)	76.0 (16.8)	81.0 (15.7)	<0.0001	71.4 (19.2)
Cigarette use, n (%)			0.2472
Current	90 (6.6)	138 (8.2)		27 (9.8)
Former	47 (3.5)	54 (3.2)		14 (5.1)
Never	1221 (89.9)	1489 (88.6)		234 (85.1)
LVEF (%), mean (SD)	65.8 (7.4)	66.5 (6.2)	0.0103	65.7 (7.7)
Atrial fibrillation/flutter rhythm on ECG, n (%)	28 (2.1)	44 (2.6)	0.3093	9 (3.3)
Medical history of:
Hypertension, n (%)	902 (66.2)	628 (37.4)	<0.0001	169 (61.4)
Hyperlipidemia, n (%)	822 (60.4)	463 (27.6)	<0.0001	124 (45.3)
Diabetes, n (%)	275 (20.3)	162 (9.7)	<0.0001	41 (22.4)
Heart failure, n (%)	39 (2.9)	29 (1.7)	0.0350	18 (6.6)
Myocardial infarction, n (%)	205 (15.0)	88 (5.2)	<0.0001	38 (13.8)
Atrial fibrillation/flutter, n (%)	113 (8.3)	96 (5.7)	0.0051	32 (11.6)
Cerebrovascular disease, n (%)	110 (8.1)	97 (5.8)	0.0122	41 (14.9)
Peripheral vascular disease, n (%)	88 (6.5)	42 (2.5)	<0.0001	26 (9.5)
Coronary revascularization, n (%)	306 (22.5)	83 (4.9)	<0.0001	38 (13.8)
PCI, n (%)	203 (14.9)	65 (3.9)	<0.0001	23 (8.4)
CABG surgery, n (%)	155 (11.4)	26 (1.6)	<0.0001	20 (7.3)
Valvular heart surgery, n (%)	41 (3.0)	19 (1.1)	0.0002	10 (3.6)
COPD, n (%)	104 (7.6)	114 (6.8)	0.3666	30 (10.9)
DVT/PE, n (%)	21 (1.5)	39 (2.3)	0.1240	8 (2.9)
Cancer, n (%)	390 (28.6)	382 (22.7)	0.0002	78 (28.4)
Medication use:
Aspirin dose >81 mg/d, n (%)	353 (25.9)			29/129 (22.5)
NSAID, n (%)	315 (23.1)	453 (27.0)	0.0154	63 (22.9)
Antihypertensive therapy n (%)	51 (3.7)	35 (2.1)	0.0060	6 (2.2)
Beta-blocker, n (%)	540 (39.6)	275 (16.4)	<0.0001	110 (40.0)
ACEi/ARB, n (%)	622 (45.6)	385 (22.9)	<0.0001	106 (38.6)
Lipid therapy, n (%)	893 (65.5)	578 (34.4)	<0.0001	131 (47.6)
Statin, n (%)	776 (56.9)	429 (25.5)	<0.0001	115 (41.8)
Non-statin, n (%)	117 (8.6)	149 (8.9)	0.7858	16 (5.8)
Diuretic, n (%)	423 (31.0)	323 (19.2)	<0.0001	82 (29.8)
Insulin, n (%)	38 (2.8)	11 (0.7)	<0.0001	12 (4.4)
Non-insulin diabetic therapy, n (%)	179 (13.1)	103 (6.1)	<0.0001	24 (8.7)
Oral anticoagulant, n (%)	52 (3.8)	102 (6.1)	0.0048	18 (6.6)
Laboratory data:
Serum creatinine (mg/dL), mean (SD)	0.95 (0.30)	0.88 (0.27)	<0.0001	1.00 (0.81)
Fasting plasma glucose (mg/dL)	110 (27)	104 (21)	<0.0001	105 (22)
Hemoglobin A1C (%), mean (SD)	5.8 (0.8)	5.7 (0.6)	<0.0001	5.8 (0.6)
Plasma lipid profile:
Total cholesterol (mg/dL), mean (SD)	175 (36)	195 (36)	<0.0001	186 (38)
LDL cholesterol (mg/dL), mean (SD)	97 (30)	112 (32)	<0.0001	104 (31)
HDL (mg/dL), mean (SD)	55 (17)	60 (18)	<0.0001	56 (17)
Triglyceride (mg/dL), mean (SD)	119 (75)	116 (68)	0.2351	131 (67)
Serum CRP (mg/L), mean (SD)	3.2 (7.4)	3.4 (7.3)	0.4580	4.2 (7.1)
Serum insulin (pmol/L), mean (SD)	83.2 (65.9)	70.5 (47.1)	<0.0001	82.4 (72.2)
Serum MCP (pg/mL), mean (SD)	383 (140)	380 (136)	0.5633	420 (136)
Serum IL-6 (pg/mL), mean (SD)	2.74 (2.94)	2.51 (2.93)	0.0395	3.38 (3.66)
Plasma P-selectin (ng/mL), mean (SD)	41.2 (13.8)	41.3 (13.5)	0.8124	40.1 (13.8)
Plasma Lp-PLA2 (ng/mL), mean (SD)	195 (52)	204 (48)	<0.0001	204 (55)
Urine 8-isoPGF2α (pg/mg creatinine), mean (SD)	1096 (599)	1149 (663)	0.0203	949 (334)
Urine albumin-creatinine ratio (mg/g), median (IQR)	61.2 (34.3, 137.8)	59.2 (34.9, 114.8)	0.2125	68.0 (42.7, 201.0)

Abbreviations: SD, standard deviation; BMI, body mass index; eGFR, estimated glomerular filtration rate; LVEF, left ventricular ejection fraction; ECG, electrocardiogram; PCI, percutaneous coronary intervention; CABG, coronary artery bypass graft; COPD, chronic obstructive pulmonary disease; DVT, deep vein thrombosis; PE, pulmonary embolus; ACEi, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; LDL, low-density lipoprotein; HDL, high-density lipoprotein; CRP, C-reactive protein; MCP, macrophage chemotactic factor; IL-6, interleukin-6; Lp-PLA₂, lipoprotein-associated phospholipase A₂; IQR, interquartile range.

Determinants of Systemic Thromboxane Generation

Previous studies in ASA users with CVD identified age, sex and oxidative stress as major independent determinants of non-platelet TXA₂ generation, with ASA dose, race, lipid therapy, left ventricular ejection fraction and renal function identified as minor independent determinants.^2,7,10,19,20 Multivariable modeling was performed to determine variables associated with TXA₂ generation in this unselected population and identify differences based on ASA use. Supplemental Table 1 shows univariate regression analyses of demographic and laboratory variables available at the time of the index examination. Multivariable modeling revealed that age, sex, oxidative stress and renal function were independently associated with TXA₂ generation irrespective of ASA use, as were cigarette use and the inflammatory markers IL-6 and P-selectin (Table 2). ASA dose, diabetes, proteinuria and atrial fibrillation/flutter were independently associated with TXB₂-M in ASA users while NSAID use, hypertension, oral anticoagulant use, COPD and HDL were associated with TXB₂-M in non-ASA users.

Table 2.

Variables independently associated with urinary TXB₂-M^a by multivariable regression analysis.

	ASA Users (N=1294)		ASA Non-users (N=1600)
Variable	Standardized Regression Coefficient	P-value	Standardized Regression Coefficient	P-value
eGFR (per mL/min/1.73m2)	0.235	<0.0001	0.200	<0.0001
Age (per year)	0.180	<0.0001	0.178	<0.0001
Female sex (versus male)	0.139	<0.0001	0.139	<0.0001
Cigarette use (versus never)
Current	0.125	<0.0001	0.046	0.0542
Former			0.061	0.0084
Urinary 8-isoPGF2α (per pg/mg creatinine)	0.069	0.0114	0.097	<0.0001
IL-6 (per pg/mL)	0.066	0.0128	0.089	0.0003
P-selectin (per pg/mL)	0.064	0.0160	0.065	0.0058
ASA dose (>81versus ≤81mg/day)	0.158	<0.0001
Urine albumin-creatinine ratio (per ln mg/g)	0.107	0.0002
Diabetes (versus none)	0.097	0.0003
Lipid therapy (versus none)	-0.063	0.0162
Atrial fibrillation/flutter (versus never)	0.054	0.0409
NSAID use (versus none)			-0.144	<0.0001
Oral anticoagulant use (versus none)			0.122	<0.0001
HDL (per mg/dL)			-0.107	<0.0001
Hypertension (versus none)			0.076	0.0021
COPD (versus none)			0.049	0.0399

Abbreviations: eGFR, estimated glomerular filtration rate; IL-6, interleukin-6; ASA, aspirin; NSAID, non-steroidal antiinflammatory drug; HDL, high density lipoprotein; COPD, chronic obstructive pulmonary disease.

^aLn pg/mg creatinine.

Association of Systemic Thromboxane Generation with Long-term Survival

Survival data was available for 3043 (99.9%) participants in whom urinary TXB₂-M was measured, 701 (23.0%) of whom died during a median observation period of 11.9 years (IQR, 10.6, 12.7 years) from the index examination. Long-term survival was significantly lower in ASA users compared to non-users (Figure 2) and was significantly associated with the degree of systemic TXA₂ generation irrespective of ASA use (Figures 3). The association between urinary TXB₂-M and all-cause mortality was present in a wide array of subgroups irrespective of ASA use (Figure 4 and Supplemental Figure 1). Urinary TXB₂-M was significantly associated with increased risk of cardiovascular death in both ASA groups, but not with stroke in either ASA group (Table 3).

Figure 2. Aspirin use and all-cause mortality risk.

All-cause mortality risk was significantly greater in the 1363 aspirin (ASA) users (blue line) compared to the 1680 ASA non-users (red line) at the time of the index examination. Number at risk over time in each group is indicated.

Figure 3. Thromboxane generation and all-cause mortality risk.

Systemic thromboxane generation was quantified by measuring urinary thromboxane B₂ metabolites (TXB₂-M). In 1363 aspirin (ASA) users (A), 1680 ASA non-users (B) and their combination (C) at the time of the index examination, all-cause mortality risk was greatest in participants with TXB₂-M in the highest quartile compared to lower quartiles. Number at risk over time in each group is indicated.

Figure 4. Thromboxane generation and all-cause mortality risk by participant characteristics.

Systemic thromboxane generation was quantified by measuring urinary thromboxane B₂ metabolites (TXB₂-M) and participant subgroups were stratified by respective quartile 4 versus quartiles 1–3 based on aspirin (ASA) use at the time of the index examination. Urinary TXB₂-M in the highest quartile was associated with increased all-cause mortality risk in a wide array of participant subgroups, irrespective of ASA use. Number of participants at risk shown in parentheses. HR=unadjusted hazard ratio; UCL=upper 95% confidence limit; LCL=lower 95% confidence limit. ASA=aspirin; BMI=body-mass index; CAD=coronary artery disease; eGFR=estimated glomerular filtration rate; LVEF=left ventricular ejection fraction.

Table 3.

Association of urinary TXB₂-M^a with unadjusted mortality risk by cause of death.

Cause of Death	ASA User (N=1363)				ASA Non-user (N=1680)				All (N=3043)b
	N (%)	HR	95% CI	P-value	N (%)	HR	95% CI	P-value	N (%)	HR	95% CI	P-value
Any	389 (28.5)	1.72	1.39–2.13	<0.0001	312 (18.6)	2.30	1.83–2.89	<0.0001	701 (23.0)	1.96	1.68–2.29	<0.0001
CVD	88 (6.5)	1.87	1.22–2.88	0.0043	45 (2.7)	3.87	2.15–6.97	<0.0001	133 (4.4)	2.41	1.71–3.39	<0.0001
Stroke	22 (1.6)	0.91	0.34, 2.45	0.8448	7 (0.4)	1.22	0.23–6.29	0.8166	29 (1.0)	0.97	0.41–2.27	0.9431
Cancer	113 (8.3)	1.28	0.85–1.91	0.2388	117 (7.0)	2.02	1.39–2.92	0.0002	230 (7.6)	1.63	1.24–2.13	0.0005
Other	136 (10.0)	1.61	1.13–2.30	0.0087	125 (7.4)	1.97	1.37–2.83	0.0002	261 (8.6)	1.78	1.38–2.29	<0.0001
Unknown	30 (2.2)	2.37	1.15–4.87	0.0192	18 (1.1)	1.17	0.42–3.25	0.7708	48 (1.6)	1.83	1.02–3.28	0.0418

Abbreviations: HR, hazard ratio; CI, confidence interval; CVD, cardiovascular disease.

^aQuartile 4 versus quartiles 1–3.

^bP-value not significant for interaction between ASA use and urinary TXB₂-M in all categories of death.

To understand the strength of association between systemic TXA₂ generation and mortality, multivariable modeling was performed to adjust for known predictors of all-cause mortality in individuals of similar age (Table 4). Urinary TXB₂-M remained significantly associated with all-cause mortality irrespective of ASA use when adjusted for age and sex (Model 2), oxidative stress and inflammation (Model 3), modifiable CVD risk factors (Model 4), or a combination of known predictors of death from heart disease, stroke, diabetes and kidney disease (Model 5). Urinary TXB₂-M was also was significantly associated with cardiovascular mortality risk when adjusted for the same variables (Models 1–4, Table 5). When adjusted for known predictors of cardiovascular death (Model 5), urinary TXB₂-M was associated with cardiovascular death in ASA non-users but not in ASA users.

Table 4.

Association of urinary TXB₂-M^a with all-cause mortality risk when adjusted for relevant predictors of survival.

Model	ASA User (N=1363)			ASA Non-user (N=1680)			All (N=3043)b
	HR	95% CI	P-value	HR	95% CI	P-value	HR	95% CI	P-value
Model 1c	1.72	1.39, 2.12	<0.0001	2.30	1.84, 2.89	<0.0001	1.96	1.68, 2.29	<0.0001
Model 2d	1.58	1.28, 1.96	<0.0001	1.68	1.34, 2.12	<0.0001	1.63	1.40, 1.91	<0.0001
Model 3e	1.67	1.34, 2.08	<0.0001	1.87	1.47, 2.38	<0.0001	1.75	1.49, 2.06	<0.0001
Model 4f	1.58	1.27, 1.97	<0.0001	2.13	1.69, 2.70	<0.0001	1.79	1.53, 2.10	<0.0001
Model 5g	1.32	1.03, 1.70	0.0288	1.60	1.22, 2.10	0.0007	1.43	1.19, 1.71	0.0001

Abbreviations: ASA, aspirin; HR, hazard ratio; CI, confidence interval; BMI, body mass index; IL-6, interleukin 6; LVEF, left ventricular ejection fraction; eGFR, estimated glomerular filtration rate.

^aQuartile 4 versus quartiles 1–3.

^bP-value not significant for interaction between ASA use and urinary TXB₂-M in all models.

^cUnadjusted.

^dAdjusted for age and sex.

^eAdjusted for 8-isoPGF_2α, IL-6 and P-selectin. (N=1299 for ASA and N=1600 for non-ASA groups.)

^fAdjusted for systolic blood pressure, total cholesterol/HDL ratio, current smoking status, BMI and hemoglobin A1_C. (N =1346 for ASA and N=1664 for non-ASA groups.)

^gAdjusted for age, sex, systolic blood pressure, total cholesterol/HDL ratio, current smoking status, atrial fibrillation, LVEF, hemoglobin A_1C, eGFR and IL-6. (N =1186 for ASA and N=1467 for non-ASA groups.)

Table 5.

Association of urinary TXB₂-M^a with cardiovascular mortality risk when adjusted for relevant predictors of survival.

Model	ASA User (N=1363)			ASA Non-user (N=1680)			All (N=3043)b
	HR	95% CI	P-value	HR	95% CI	P-value	HR	95% CI	P-value
Model 1c	1.87	1.22, 2.88	0.0043	3.87	2.15, 6.97	<0.0001	2.41	1.71, 3.39	<0.0001
Model 2d	1.69	1.08, 2.64	0.0219	2.69	1.46, 4.96	0.0015	1.93	1.35, 2.76	0.0003
Model 3e	1.68	1.06, 2.65	0.0259	3.69	1.95, 6.96	<0.0001	2.11	1.46, 3.06	<0.0001
Model 4f	1.65	1.02, 2.65	0.0397	3.56	1.92, 6.62	<0.0001	2.13	1.47, 3.09	<0.0001
Model 5g	1.36	0.77, 2.38	0.2898	2.82	1.39, 5.66	0.0041	1.70	1.10, 2.62	0.0163

Abbreviations: ASA, aspirin; HR, hazard ratio; CI, confidence interval; BMI, body mass index; IL-6, interleukin 6; LVEF, left ventricular ejection fraction; eGFR, estimated glomerular filtration rate.

^aQuartile 4 versus quartiles 1–3.

^bP-value not significant for interaction between ASA use and urinary TXB₂-M in all models.

^cUnadjusted.

^dAdjusted for age and sex.

^eAdjusted for 8-isoPGF_2α, IL-6 and P-selectin. (N=1299 for ASA and N=1600 for non-ASA groups.)

^fAdjusted for systolic blood pressure, total cholesterol/HDL ratio, current smoking status, BMI and hemoglobin A1_C. (N =1346 for ASA and N=1664 for non-ASA groups.)

^gAdjusted for age, sex, systolic blood pressure, total cholesterol/HDL ratio, current smoking status, LVEF, hemoglobin A_1C, eGFR and IL-6. (N =1186 for ASA and N=1467 for non-ASA groups.)

Given that ASA was frequently used for primary prevention during the time-period of the index examinations (2005–2008), we investigated the association of urinary TXB₂-M with all-cause and cardiovascular mortality in the 2353 participants without established CVD. The risk of all-cause but not cardiovascular mortality was greater in the 898 (38.2%) participants taking ASA compared to the 1455 (61.8%) not taking ASA at the index examination (Figure 5). Survival plots of ASA users and non-users stratified by urinary TXB₂-M quartile grouping are shown in Figure 6. Analyses accounting for the interaction between urinary TXB₂-M and ASA use revealed the former to be robustly associated with all-cause and cardiovascular mortality in participants not taking ASA but not in those taking ASA at the index examination, associations that remained significant after adjustment for modifiable CVD risk factors (Figure 7A). ASA use was associated with increased risk of all-cause and cardiovascular mortality in participants with low levels (quartiles 1–3) but not with high levels (quartile 4) of urinary TXB₂-M (Figure 7B).

Figure 5. Aspirin use and mortality risk in participants without cardiovascular disease.

All-cause mortality (A) but not cardiovascular mortality (B) risk was greater in the 898 participants taking ASA (blue line) compared to the 1455 not taking ASA (red line) at the time of the index examination. Number at risk over time in each group is indicated.

Figure 6. Thromboxane generation and mortality risk in participants without cardiovascular disease.

Systemic thromboxane generation was quantified by measuring urinary thromboxane B₂ metabolites (TXB₂-M) in 2353 participant stratified by aspirin (ASA) use and without cardiovascular disease at the index examination. Urinary TXB₂-M in the highest quartile (blue line) was associated with grater all-cause mortality risk compared to lower quartiles (red line) in 898 ASA users (A) and 1455 ASA non-users (B) and was associated with increased cardiovascular mortality risk in ASA non-users (D) but not in ASA users (C). Number at risk over time in each group is indicated.

Figure 7. Interaction of aspirin use and thromboxane generation with mortality risk.

Systemic thromboxane generation was quantified by measuring urinary thromboxane B₂ metabolites (TXB₂-M) in 2353 participant without established cardiovascular disease stratified by aspirin (ASA) use at the index examination. Interaction analyses revealed all-cause and cardiovascular mortality risk was greater in ASA non-users, but not ASA users with urinary TXB₂-M in the highest quartile (A) as well as in ASA users with urinary TXB₂-M in the lowest quartiles (B), even after adjustment for modifiable cardiovascular risk factors. Numbers of participants at risk shown in parentheses. HR=unadjusted hazard ratio; UCL=upper 95% confidence limit; LCL=lower 95% confidence limit. ASA=aspirin; BMI=body mass index. ^aUnadjusted. ^bAdjusted for systolic blood pressure, total cholesterol/HDL ratio, smoking status, BMI and hemoglobin A1C.

DISCUSSION

The major findings of this study (Central Illustration) include: 1) Systemic TXA₂ generation is significantly associated with long-term all-cause and cardiovascular mortality in a large unselected population irrespective of ASA use and remains so after adjustment for known risk factors for CVD and mortality, and; 2) In participants without CVD, systemic TXA₂ generation is associated with all-cause and cardiovascular mortality after adjustment for modifiable CVD risk factors in ASA non-users, but not in ASA users, while ASA use is associated with increased mortality in those with lower levels, but not in those with high levels, of systemic TXA₂ generation.

Central Illustration. Schematic of systemic thromboxane generation and association with long-term mortality.

Aspirin (ASA) effectively suppresses thromboxane A₂ (TXA₂) generation in platelets but not in endothelial cells under oxidative stress or activated inflammatory cells. TXA₂ generation and signaling through cellular thromboxane-prostanoid receptors has been implicated in multiple disease processes and increased systemic TXA₂ generation, as evidenced by elevated levels of stable urinary thromboxane B₂ metabolites (TXB₂-M), is associated with reduced long-term survival irrespective of ASA use. In individuals without established cardiovascular disease, measurement of urinary TXB₂-M may be useful to identify those who would benefit from or be harmed by ASA therapy for primary prevention.

TXA₂ generation is most recognizably associated with platelet activation, both as an intracellular amplifier of signaling in response to multiple physiologic agonists and as a potent extracellular agonist via binding to surface thromboxane-prostanoid receptors (TPr).¹ While persistent systemic TXA₂ generation despite ASA use was initially thought to be due to incomplete ASA-mediated suppression of platelet COX-1, it is now appreciated that daily ASA doses of 81–325 mg are effective at inhibiting platelet TXA₂ generation in the vast majority of individuals and that residual TXA₂ generation in compliant ASA users originates predominantly from non-platelet tissues.^2–5,21,22 The circulating half-life of ASA is ~20 minutes with negligible blood levels 6–8 hours after ingestion of a single 81mg dose.²³ Unlike platelets that lack capacity to regenerate COX-1 irreversibly acetylated by ASA, nucleated cells can produce TXA₂ following regeneration of COX-1 or by an inducible COX-2-mediated pathway that is not effectively inhibited by standard doses of ASA.²⁴ Though standard ASA therapy does not completely suppress TXA₂ generation in non-platelet tissue, the inverse association between TXB₂-M and ASA dose observed in ours and other studies suggest that there is at least some degree of suppression.^2,7

Prior studies identifying an association of systemic TXA₂ generation with adverse outcome were mostly performed in ASA users with established CAD.^6–10 To our knowledge, ours is the largest study to measure systemic TXA₂ generation in an unselected population and the first to identify strong associations with all-cause and cardiovascular mortality risk in both ASA users and non-users. Importantly, the association of systemic TXA₂ generation with mortality remained strong after adjustment for age, sex and oxidative stress, previously identified risk factors for non-platelet TXA₂ generation, as well as markers of inflammation, cardiovascular disease, renal disease and diabetes that are recognized predictors of survival in middle-aged and older individuals. Association of systemic TXA₂ generation with mortality risk was further observed in a diverse array of subgroups irrespective of ASA use, including those without established CVD at the time of the index examination.

Particularly intriguing was the observed interaction between ASA use and systemic TXA₂ generation in participants without CVD and their respective association with mortality risk. ASA use alters systemic TXA₂ generation, an effect that was accounted for by using quartile groupings of urinary TXB₂-M based on ASA use. While we did not find significant interactions between ASA use and urinary TXB₂-M in the full study cohort, we did observe a significant interaction in the subgroup of participants without CVD. Accounting for that interaction, the mortality risk associated with urinary TXB₂-M and ASA use were interdependent and remained significant after adjustment for modifiable CVD risk factors: Elevated urinary TXB₂-M was associated with increased mortality risk only in ASA non-users and ASA use was associated with increased mortality risk only in those with lower levels of urinary TXB₂-M. While the indications for ASA use in individuals without CVD have evolved substantially over the last few years, uncertainty remains regarding its potential benefit in certain high-risk groups.²⁵ Our data raise the possibility that measurement of urine TXB₂-M in individuals without established CVD might potentially help identify those who could benefit or be harmed by ASA therapy for primary prevention, a premise that warrants evaluation in future studies.

Major in vivo sources of non-platelet TXA₂ generation have yet to be identified but likely differ among various patient populations. Like previous studies, we found a strong association of urinary TXB₂-M irrespective of ASA use with oxidative stress as measured by urinary 8-iso-PGF_2α.^2,19,26,27 Isoprostanes, such as 8-iso-PGF_2α, are formed non-enzymatically from arachidonic acid under conditions of oxidative stress. 8-iso-PGF_2α is not only a marker of oxidative stress but it is also a biologically active stimulatory ligand of TPr, though of lower affinity than TXA₂ ²⁸ Oxidative stress is intimately involved in the pathobiology of a myriad of human diseases and is the sine qua non of endothelial dysfunction.²⁹ In vitro studies by our group and others revealed that macro- and microvascular endothelial cells exposed to oxidative stress generate both TXA₂ and 8-isoPGF_2α, and that 8-isoPGF_2α can directly stimulate TXA₂ generation via activation of the TPr.^2,30,31 We also observed an independent association between urinary TXA₂-M and IL-6, an inflammatory cytokine produced by activated monoctes/macrophages and activated endothelial cells that has been associated with adverse cardiovascular outcome.³² TXA₂ is generated in monocytes via an inducible COX-2 pathway where it influences acquired immunity.³³ In a study of cardiac bypass graft surgery patients in whom platelet TXA₂ was documented to be completely suppressed, we found non-platelet TXA₂ generation in the early post-operative period to be 3-fold greater and predominantly associated with the degree of inflammation compared to 6 months later when the intense post-operative inflammation had subsided and urinary TXB₂-M was more strongly associated with oxidative stress.⁹ In a study of hypertensive individuals not taking ASA the administration of a selective COX-2 inhibitor did not suppress systemic TXA₂ generation, suggesting that mononuclear cells may only be a significant source of systemic TXA₂ generation in situations of intense inflammation.³⁴

While TXA₂ generation in platelets is clearly a key process in the development of atherothrombosis, it is not clear how TXA₂ generation in non-platelet tissue adversely influences outcome. Evidence does not suggest that TXA₂ generated in non-platelet tissue causes pathologic thrombosis by stimulating platelet activation. The integrin P-selectin is highly expressed in both endothelial cells and platelets where it mediates adhesion of neutrophils and platelets to the endothelium and its soluble form is thought to predominantly originate from activated platelets.³⁵ Median plasma P-selectin levels in our cohort did not differ between ASA users and non-users, suggesting similar degrees of platelet reactivity in both groups, despite a more than three-fold higher median urinary TXB₂-M in the latter. Though non-platelet TXA₂ generation was associated with an increased risk of early graft thrombosis and death after coronary artery bypass graft surgery, that risk was independent of measures of platelet reactivity and the risk of adverse outcome associated with increased systemic TXA₂ generation in ASA users with CVD was not attenuated by the addition of the ADP-receptor antagonist clopidogrel.^7,8 TPr is expressed to variable degrees in nearly all tissues of the body and its activation is known to elicits a myriad of biological effects in non-platelet tissue, including vasoconstriction, atherogenesis, fibrosis, inflammation, immune modulation and cancer progression.^36–38 Increased systemic TXA₂ generation has been observed in several conditions where these non-platelet TPr signaling effects are known to play prominent pathologic roles. For example, TXA₂ is a potent systemic/intrarenal vasoconstrictor and urinary TXB₂-M has been shown to be elevated in patients with heart failure and proportional to its severity.^39,40 TPr activation is arrhythmogenic in cardiomyocytes by altering intracellular calcium signaling and urinary TXB₂-M is elevated in atrial fibrillation where it associates with adverse outcomes.^41–43 TPr expression is increase in a variety of cancer types and increased urine TXB₂-M has been observed in patients with bladder, prostate, breast and colorectal cancer.^44–47 Our study adds to the body of evidence linking TXA₂ generation and TPr signaling with malignancy by showing a strong association of systemic TXA₂ generation to cancer mortality.

The robust association between systemic TXA₂ generation and mortality observed in our study provides a rationale to investigate whether inhibiting non-platelet TXA₂ generation or blocking its biological effects could improve survival in patients with CVD and potentially other diseases. Extended-release ASA could theoretically achieve more persistent inhibition of non-platelet COX-1 activity than standard once daily ASA therapy. A micronized formulation that leads to sustained blood levels of acetylsalicylic acid has been shown to effectively reduce systemic TXA₂ generation in healthy individuals, though it has not been evaluated in populations with high degrees of non-platelet TXA₂ generation.²³ Thromboxane synthase inhibitors and TPr antagonists also have the potential to suppress both platelet and non-platelet TXA₂ generation without impacting other prostaglandin pathways. While initially developed as antiplatelet agents, TPr antagonists are currently being evaluated in human clinical trials to improve endothelial function, reduce cardiac and pulmonary fibrosis as well as suppress cancer metastases.^48–51

Study Limitations

There are several acknowledged limitations of our study. ASA use was determined at the time of the index examination and used to phenotypically stratify participants into those whose urine TXB₂-M reflects predominantly non-platelet-TXA₂ generation (ASA users) and those in whom it reflects both platelet and non-platelet TXA₂ generation (ASA non-users) at a given point in time. We could not account for changes in use of ASA or other medications over time that could potentially impact survival. Changes in the magnitude of systemic TXA₂ generation overtime could also potentially alter risk. This study and others revealed that urinary TXB₂-M exhibits an inverse relationship to renal function.² This is unlikely to be a significant confounder because it is small in magnitude (e.g., for every 10 mL/min/1.73m² decrease in eGRF urinary TXB₂-M increases by 1.1 pg/mg creatinine) and would tend to diminish the correlation of urinary TXB₂-M to death. The Offspring cohort of the Framingham Heart Study enrolled almost exclusively white individuals of European ancestry. While 298 participants enrolled in the minority Omni cohort were included in our analysis, non-white participants constituted <10% of the entire study population. Although previous studies have indicated that non-whites have higher levels of non-platelet TXA₂ generation than whites, we could not adequately investigate the impact of race on outcome due to this underrepresentation of minorities in the study cohort.

Conclusions

Systemic TXA₂ generation is an independent risk factor for long-term all-cause and cardiovascular mortality in an unselected cohort of older individuals irrespective of ASA use. Measuring TXA₂ generation in individuals without CVD may be useful in guiding the aggressiveness of primary prevention strategies, including use of ASA. These findings provide a rationale for investigating whether inhibiting TXA₂ generation or blocking its effects could impact survival in individuals with CVD and potentially other conditions.

PERSPECTIVES.

Competency in Medical Knowledge:

Aspirin reduces cardiovascular risk by inhibiting thromboxane generation in platelets, though it is much less effective at inhibiting thromboxane generation in non-platelet tissues. Systemic thromboxane generation is robustly associated with all-cause and cardiovascular mortality in a general population of both aspirin users and non-users.

Translational Outlook 1:

Measurement of systemic thromboxane generation may help to better risk stratify individuals irrespective of aspirin use, especially in those without established cardiovascular disease who may benefit from aggressive primary prevention.

Translational Outlook 2:

Inhibiting or blocking the effects of non-platelet thromboxane generation may be a novel strategy to improve outcomes in individuals with high levels of systemic thromboxane generation.

ABREVIATIONS

ASA

aspirin

BMI

body mass index

COPD

chronic obstructive pulmonary disease

COX

cyclooxygenase

CVD

cardiovascular disease

eGFR

estimated glomerular filtration rate

FHS

Framingham Heart Study

LVEF

left ventricular ejection fraction

TPr

thromboxane-prostanoid receptor

TXA₂

thromboxane A₂

TXAS

thromboxane synthase

TXB₂-M

thromboxane B₂-metabolites