Usefulness of Serum Unbound Free Fatty Acid Levels to Predict Death Early in Patients with ST Segment Elevation Myocardial Infarction[From the TIMI II Trial]

Abstract

Circulating total free fatty acids (FFA) are elevated early in myocardial infarction (MI) and are associated with an increase in mortality. We investigated the association of serum unbound free fatty acids (FFA_u) levels with mortality,in patients presenting with ST elevation myocardial infarction (STEMI) in the Thrombolysis in Myocardial Infarction (TIMI) II trial.TIMI II enrolled patients within 4 hours of chest pain. Patients were treated with recombinant tissue plasminogen activator within 1 hour of enrollment. The concentration of FFA_u was evaluated in serum samplesfrom 1834 patients obtained at baseline, before therapy.FFA_u was an independent risk factor for death as early as one day of hospitalization and continued to be an independent risk factor for the more than 3·8 years of follow up. When adjusted for other cardiovascular risk factors FFA_u levels in the fourth as compared to the first quartile remained an independent risk factor for death due to MI (hazard ratio, 5.0; 95 % confidence interval, 1.9-13.0), to all cardiac death (hazard ratio, 2.4; confidence interval, 1.3-4.4) and to all cause death (hazard ratio, 1.9, confidence interval, 1.2-3.1).Females were twice as likely to be in the upper two FFA_u quartiles and had approximately twice the rate of death as males. In conclusion, increased levels of FFA_u are one of the earliest molecular biomarkers of mortality in STEMI and are independent of other risk factors known to affect outcomes in STEMI.

Plasma free fatty acid (FFA) levels are elevated early after acute myocardial infarction (MI) and correlate with increased rates of arrhythmias and mortality, particularly within the first 12 hours.^1-4 The MI-associated FFA increase occurs primarily through catecholamine activation of adipose tissue lipolysis rather than FFA release from the ischemic cardiac tissue.⁴ Although most plasma FFA is bound to albumin, a small fraction (< 10⁻⁴) is unbound FFA (FFA_u). FFA_u levels increase exponentially with increasing ratio of total FFA to albumin and thus FFA_u are more sensitive to physiologic changes than total FFA.^5;6FFA_ulevels increase rapidly, within 30 minutes of cardiac ischemia induced by balloon angioplasty.^7;8We investigated whether FFA_ulevels from patients in the Thrombolysis in Myocardial Infarction (TIMI)Phase II trial of STEMI provide an independent assessment of risk for poor outcomes at timesas early as 24 hours after symptoms.

METHODS

The TIMI II trial treated 3262 patients,who presented within 4 hours of STEMIonset,with intravenous recombinant tissue plasminogen activator plus heparin.⁹After recombinant tissue plasminogen activatortherapy patients were randomly assigned to either a percutaneous coronary interventionor a conservative strategyin which only patients exhibiting ischemia (13%) received percutaneous coronary intervention. TIMI investigators recorded demographics, medical history and outcomes over a follow up period of 3.8 years. A Limited Access DataSetof patient parameters was available for this study.

Measurements were performed on serum samples (baseline) drawn prior to recombinant tissue plasminogen activator and heparin therapy.¹⁰Serum wascollected in STEMI patients by the TIMI investigators ¹¹ and maintained at −70°C by the National Heart Lung Blood Institute blood specimen repository. A subset of these specimens from 2500 patients was provided by the National Heart Lung Blood Institute. The linkage between patient information and blood specimen tube identification was maintained separately by the Maryland Medical Research Institute. After our results were deposited with the National Heart Lung Blood Institute, the linkage was un-blinded. Complete patient information from the Limited Access Data Setwas available for 1834 patients of the 2500 baseline blood specimens and results reported here are for the 1834 patients. This study complies with the Declaration of Helsinki and was approved by the Institutional Review Board committee of the Torrey Pines Institute for Molecular Studies.

Measurements of serum FFA_u concentrations were performed using the fluorescent probe ADIFAB2 (FFA Sciences) as described previously in cardiac ischemia and MI patients^8;12 but modified for 96-well plate fluorometry using a Flurolog 3 spectrofluorometer with a MicroMax plate reader (J.Y Horiba). Serumsamples were diluted to 1% (v/v) in 200 L of measuring buffer¹³ in 96-well plates. Fluorescence was measured after addingADIFAB2 (1.5 mol/L) and the intensities were used to determine the ratio (R) of the fluorescence intensitiesat 550 to 457 nm,with background subtracted. Sample FFA_u concentrations (nmol/L) were calculated using FFA_u=227(R-R_o)/(0.925 – R), where R_o is the ADIFAB2 fluorescence ratio in the absence of FFA and the numeric factors were determined as described previously ⁶. All samples were measured in duplicate, yielding an average CV of 6.5 %.

FFA_u quartiles were determined using all baseline FFA_u values (Table 1). Correlations between FFA_u and patient baseline characteristics were carried out usingallcardiovascular confounders available in the limited data set. In TIMI II, patients with renal disorders were excluded, Killip class was not recorded and creatine kinase was the only cardiac biomarker measured. Outcomes of death due to STEMI, to cardiac causes and to all causes were determined through committee adjudication by the TIMI investigators. Statistical analyses were performed using XLSTAT (Addinsoft, New York), p values ≤ 0.05 were considered significant.

Table 1.

Baseline parameters for TIMI II patients as a function of unbound free fatty acid quartiles.

Number in quartiles	Quartile1 458	Quartile2 458	Quartile3 459	Quartile4 459	p Value
Unbound free fatty acids (nmol/L)	1.9 (0.09 – 2.6)	3.2 (2.6 – 3.9)	4.9 (3.9 – 6.4)	10.2 (6.46 – 523)
Age (years)	53.9	53.8	55.3	55.4	0.008
Men	398(87%)	389(85%)	344(75%)	372(81%)	<0.0001
White	398(87%)	398(87%)	409(89%)	398(87%)	0.62
Body Mass Index (kg/m2)	27 (15-40)	27 (18-46)	27 (16-54)	27 (17-54)	0.55
Diastolic blood pressure (mm Hg)	81 (42 -130)	81 (50 -110)	80.2 (40-120)	78.2 (40-118)	0.053
Systolic blood pressure (mm Hg)	130 (56 -180)	130 (60-183)	129 (80-210)	126 (50 -190)	0.044
Creatine Kinase (IU/L)	105 (5 -1760)	104 (12 -3590)	94 (11 -4422)	99 (11-3654)	0.080
Diabetes mellitus	41(9%)	37(8%)	73(16%)	78(17%)	<0.0001
Prior MI	60(13%)	64(14%)	69(15%)	73(16%)	0.50
Prior hypertension	174(38%)	156(34%)	188(41%)	179(39%)	0.20
β blocker within 24 hours	87(19%)	87(19%)	78(17%)	73(16%)	0.68

Values for ordinal parameters are percent and p values were calculated by χ². Continuous parameters are median values and intra-quartile ranges. p values were determined by the Kruskal-Wallis test. Ages were grouped to protect confidentiality and are proportional to the grouped mean for each quartile. Baseline creatine kinase levels were elevated above the upper limit of normal in fewer than 20 % of patients. Mean time from chest pain symptom to treatment initiation was virtually identical (2.6 h) for each FFA_u quartile (data not shown). Demographic and clinical parameters in this table are representative of the entire TIMI II population.¹⁰

RESULTS

Baseline FFA_u levels were measured in serum samples collected within 4 hours of initial symptoms. FFA_u concentrations ranged from < 1 nmol/L to > 500 nmol/L (Table 1). This range is larger than we reported previously for a cohort of nominally healthy subjects whose range was 0.6 to 4.5 nmol/L with a mean value of 1.5 nmol/L.¹²Correlations of FFA_u quartiles with all cardiovascular risk factors available from the limited access data set and for which patientswith the indicated risk factors comprised at least 10% of all patients are shown in Table 1.

Of the 1834 patients, 187 died from all causes, 125 from cardiac causes and 76 from MI (Table 2). A positive correlation of death from MI with FFA_u quartile was present as early as one day following enrollment and peaked at about 30 days post enrollment. Because most deaths within 30 days were due to MI, cardiac deaths and deaths from all causes also correlated strongly with baseline FFA_u. Peak (within 8 hours, Table 2), but not baseline (Table 1) creatine kinase levels,were correlated with FFA_u.

Table 2.

Death and outcomes according to baseline unbound free fatty acid quartiles

Median unbound free fatty acid (nmol/L)	Quartile1 1.9	Quartile2 3.2	Quartile3 4.9	Quartile4 10.2	Total	p Value	p Trend
Death due to:
Myocardial Infarction at 1 day	3	7	11	12	33 (1.8%)	0.1	0.015
Myocardial Infarction at 7 days	3	7	19	21	50 (2.7%)	0.0002	<0.0001
Myocardial Infarction at 30 days	4	10	21	29	64 (3.5%)	<0.0001	<0.0001
Myocardial Infarction at 3.8 years	6	14	24	32	76 (4.1%)	<0.0001	<0.0001
All cardiac causes at 1 day	3	7	11	12	33 (1.8%)	0.1	0.015
All cardiac causes at 7 days	3	7	19	21	50 (2.7%)	0.0002	<0.0001
All cardiac causes at 30 days	5	11	21)	29	66 (3.6%)	<0.0001	<0.0001
All cardiac causes at 3.8 years	16	27	38	44	125 (6.9%)	0.0006	<0.0001
All causes at 1 day	3	8	13	13	37 (2.0%)	0.056	0.0095
All causes at 7 days	3	9	24	23	59 (3.2%)	<0.0001	<0.0001
Al causes at 30 days	5	13	26	34	78 (4.3%)	<0.0001	<0.0001
All causes at 3.8 years	29	38	57	63	187 (10%)	0.0004	<0.0001
Non cardiovascular causes at 3.8 years*	13	9	13	15	50 (2.7%)	0.67	0.530
Other outcomes
Peak Creatine Kinase (8 hours) (IU/L)	1628	1957	2039	2156		<0.0001

All other outcomes are numbers of patients. The number of deaths from all causes is equal to all cause cardiac plus non-cardiac plus hemorrhage. Values in parenthesis are % of total number of patients (1834). p values were determined by the χ² test for all but peak creatine kinase for which the Kruskal-Wallis test was used. p trend values were calculated using the Cochran-Armitage trend test.

^*An additional 12 patients died due to hemorrhage over the 3·8 year follow up period (4 at day 1). Adding all cardiac plus non-cardiovascular deaths, plus one death without cause sums to 187.

Relative to the total TIMI II population FFA_u levels correlated positively with female gender, age, and diabetes. Approximately twice as many females and diabetics were in the upper two FFA_u quartiles (Table 1). Deaths due to MI in non-diabetic females and males as well as in diabetics increased with increasing FFA_u quartile (Table 3). This correlation reached significance in non-diabetic females and males, but not obtained in diabetics. The lack of correlation in diabetics is likely due to small numbers and to two deaths within 5 hours in the first quartile, but no additional deaths in the first quartile over the following 3.8 years. The death rate for females and diabetics was almost 2-fold > for non-diabetic males,except for day 1, mostly due to larger death rates at 30 days in the 3^rd and 4^th FFA_u quartiles. Using subgroup interaction analysis death rates in Q3+Q4 were significantly (p=0.05) higher in non-diabetic females thanmales but not (p=0.07) inall diabetics versus non-diabetics¹⁴.

Table 3.

Myocardial infarction deaths in females, males and diabetics relative to baseline unbound free fatty acids.

Unbound free fatty acid Quartiles	Quartile1	Quartile2	Quartile3	Quartile4	Total	p Value	p Trend
Non Diabetic Females (n=283):
By day 1	0	1	1	3	5(1.8%)	0.02	0.10
By 7 days	0	1	5	6	12(4.2%)	0.08	0.009
By 30 days	0	2	5	8	15(5.3%)	0.04	0.004
By 3.8 years	1	2	6	8	17(6.0%)	0.1	0.06
Non Diabetic Males (n=1322):
By day 1	1	6	7	6	20(1.5%)	0.9	<0.022
By 7 days	1	6	9	10	26(2.0%)	<0.014	<0.0018
By 30 days	2	7	11	14	34(2.6%)	<0.003	<0.0003
By 3.8 years	3	10	12	17	42(3.2%)	<0.002	<0.0002
Diabetics (n=229):
By day 1	2	0	3	3	8(3.5%)	0.11	0.86
By 7 days	2	0	5	5	12(5.2%)	0.44	0.37
By 30 days	2	1	5	7	15(6.6%)	0.57	0.23
By 3.8 years	2	2	6	7	17(7.4%)	0.79	0.32

Data are numbers of deaths and percent of each subgroup (n). p values were determined by χ² and p trend by the Cochran-Armitage trend test.

Kaplan-Meier survival curves for deaths due to STEMI, to all cardiac causes and to all causes reveal striking dependencies on FFA_u quartiles (Figure 1). Log-rank tests indicate that the risk of death for all three categories increased significantly (p < 0.001) with increasing FFA_u quartile. For MI almost all deaths occurred within the first 30 days. For deaths from all causes the survival curve reveals a similar slope after 60 daysfor all 4 quartiles, consistent with a lack of correlation with FFA_u for non-cardiac mortality (Table 2). Cox proportional hazard modeling of the three sets of survival curves reveals hazard ratios that increase significantly with FFA_u quartile relative to the first quartile and are relatively unaffected by other risk factors (Table 4). The unadjusted hazard ratios for MI deaths increased from 2.3 for Q2 to 5.6 for Q4 and the corresponding adjusted hazard ratios were 2.9 and 5, respectively. Except for the Q1 to Q2 increase, the hazard ratios increases with FFA_u quartile were significant (p ≤ 0.03) for all unadjusted and adjusted analyses.

Figure 1. Kaplan-Meier survival curves for death by quartiles of baseline FFA_u.

Survival curves for Q1 to Q4 for death due to A) MI, B) all cardiac causes and C) all causes. Log-rank probabilities were < 0·001.

Table 4.

Cox proportional hazard ratios for deaths relative to the first unbound free fatty acid quartile

	Quartilea	Hazard Ratio	CI (95%)	p Value
Myocardial Infarction Deaths
Unadjusted	Q2	2.3	(0.9 – 5.9)	0.094
	Q3	4	(1.6 – 9.8)	0.002
	Q4	5.6	(2.3 – 13)	< 0.001
Adjusted	Q2	2. 9	(1.0 – 8.2)	0.043
	Q3	4.2	(1.6 – 11)	0.002
	Q4	5	(1.9 – 13)	0.001
All Cause Cardiac Deaths
Unadjusted	Q2	1.7	(0.9 – 3.1)	0.103
	Q3	2.4	(1.3 – 4.3)	0.003
	Q4	2.8	(1.6 – 5.0)	0.000
Adjusted	Q2	1.7	(0·9 – 3.2)	0.108
	Q3	2.1	(1.2 – 4.0)	0.015
	Q4	2.4	(1.3 – 4.4)	0.004
All Cause Deaths
Unadjusted	Q2	1.3	(0.8 – 2.1)	0.324
	Q3	1.9	(1.2 – 3.0)	0.004
	Q4	2.2	(1.4 – 3.4)	0.001
Adjusted	Q2	1.3	(0.8 – 2.1)	0.374
	Q3	1.7	(1.0 – 2.7)	0.032
	Q4	1.9	(1.2 – 3.1)	0.004

^aCox proportional hazard model generated hazard ratios and corresponding 95 % confidence interval (CI) and Wald's χ² probabilities. The model was adjusted for age, gender, race, body mass index, diastolic and systolic blood pressure at baseline, history of diabetes, MI and hypertension and use of β blockers within 24 hours of enrollment. This choice of factors was dictated by the limited access data set, known cardiac risk factors associated with more than 10 % of the 1834 patients and previous observed associations with free fatty acids.

DISCUSSION

Oliver and Opie recognized more than 40 years ago that serum levels of FFA increase rapidly following MI and that the elevated FFA levels may, by inducing arrhythmias, contribute to sudden cardiac death and death in MI.^1-4Moreover, these studies found that early death in MI was highly correlated with increasing total FFA levels ¹. Total FFA levels were also correlated with the long term (6.9 years) increased risk of sudden cardiac death inthe LURIC study.¹⁵Our results are consistent with and additive to LURIC.¹⁵Both the present and LURIC studies found strong baseline correlations between FFA_u or total FFA levels and female gender or a history of diabetes (Table 1).Hazard ratios for MI deaths in TIMI II were larger (≥3 fold) and less sensitive to other risk factors than in the LURIC study. Presumably, this increase in sensitivity is a reflection of the acute presentation of STEMI in TIMI II and the high degree of FFA_u sensitivity to cardiac ischemia ^7;8.

The lower survival of females and diabetics as compared to non-diabetic men (Table 3) is consistent with other studies in STEMI patients.^16;17In TIMI II females had higher FFA_u levels than men; median FFA_u for non-diabetic men and females were 3·7 nmol/L and 4.5 nmol/L, respectively.The higher FFA_u levels for femalesin TIMI II might reflect a gender specific difference in the response to STEMI or might be a related to the older (7 years) age of females in TIMI II, given that FFA_u increase with age (Table 1). However, the adjusted (see Table 4 legend) hazard ratio for death due to MI for females in Q4 versus Q1 is 9·7 (95% CI: 0.971 to 100) with p = 0.053. These results, and the higher death rates in Q3+Q4 for females, raise the possibility that FFA_u may be a more potent independent risk factor for females than for males.

Although these results demonstrate that FFA_u correlate with early death after STEMI, causation is unproven. The increase in circulating FFA_u may simply reflect the ischemia-induced activation of adipose lipolysis in proportion to the degree of ischemia and is probably largely generated from adipose tissue lipolysis stimulated by an ischemia-mediated increase in catecholamine levels.^3;4In contrast, evidence for FFA having a causal role are studies in nonischemic animals in which increasing circulating FFA adversely affect myocardial metabolism, stimulate insulin resistance, induce arrhythmias and increase cardiac enzyme release.^18-21Consistent with a causal role for FFA_uin TIMI II are the increase in peak (8 hours) but not baselinecreatine kinase levels with increasing baseline FFA_u(Table 3).

If FFA adversely affectmyocardial function, reducing FFA_u levels at times early after the ischemic event may reduce deaths and/or arrhythmias in MI.^3;4Infusions of glucose-insulin-potassium have been, shown to be protective in the dog ²², were used to treat STEMI patients andreduced circulating total FFA acutely in STEMI patients.²³Although the effect on outcome in STEMI patients has been mixed, it has been suggested that GIK treatment would be most effective if given at the earliest possible time.^24;25 This concept was implemented in the IMMEDIATE trial by treating STEMI patients with glucose-insulin-potassium in pre-hospital emergency medical settings.^25;26The IMMEDIATE trial revealed that glucose-insulin-potassium significantly reduced FFA levels, reduced cardiac arrest plus in-hospital mortality and reduce infarct size as compared to patients not treated with glucose-insulin-potassium. Conceivably, therapeutic efficacy of such interventions, using glucose-insulin-potassium or inhibitors of lipolysis, would be most evident in those MI patients who present with the highest levels of FFA_u; for which further studies are necessary.

Evidence-based guidelines currently recommend the rapid application of re-perfusion therapy with primacy given to percutaneous coronary artery intervention, with limited use of tissue plasminogen activator,the therapy in TIMI II. Nevertheless,the30 day death rate in TIMI II (4.2%) was not significantly differentthan the 5.4% in contemporary STEMI patients.¹⁶Adjustments for confounders that might affect hazard ratios for FFA_uwere limited because important prognostic factors²⁷,includingKillip class, baseline heart rate, number of diseased vessels, ejection fraction,smokingand troponin, were not available or not recorded in the limited access data set. ACE inhibitors and statins were first approved for use in 1981 and 1987, respectively. TIMI II enrollment wasbetween 1986 and 1988 and therefore most patients were unlikely to have been treated with these medications. Catecholamines are associated, weakly, with mortality in STEMI ²⁸ but were not measured. The beneficial effects of glucose-insulin-potassium suggest that FFA not catecholamines areimportant contributors to mortality in STEMI.TIMI II protocols did not anticipate determination of FFA levels and therefore blood samples may have had higher ex vivo than in vivo FFA as a consequence of lipoprotein lipolysis.²⁹However, this ex vivo effect would have obscured rather than enhanced the observed FFA_u correlations with outcome. The determination of FFA_u concentrations requires knowledge of the relative distribution of the different FFA_u present in serum and was estimated from the distribution of total FFA as described in references.^6;13Differences in the FFA_u distributions such as those that may occur in acute coronary syndromes³⁰ are not expected to alter the levels of total FFA_u significantly.⁶The study was borderline under powered for female patients, suggesting the need for further studies.