Previous clinical and experimental studies suggested that lipid-reducing drugs such as niacin and statins may affect mitochondrial and muscle function (1-3). Statins, which are commonly used drugs for treating hypercholesterolemia, competitively inhibit 3-hydroxy-3-methyl-glutaryl coenzyme A (HMG-CoA) reductase, the rate-limiting enzyme in cellular cholesterol synthesis. Important nonsterol compounds, such as ubiquinone (known as coenzyme Q10), also are derived from the same synthetic pathway (4). Accordingly, inhibition of HMG-CoA reductase may cause, in addition to reduced cholesterol synthesis, a decrease in the production of ubiquinone, at least in vitro (5,6). A main function of ubiquinone in the cell is to serve as a lipid-soluble carrier in the membrane-bound electron transport chains of the mitochondria. Therefore it has been argued that a decrease in serum or tissue ubiquinone levels during statin treatment would lead to a reduction in its bioavailability (1) and possibly to tissue damage through mitochondrial dysfunction and reduced mitochondrial generation of high-energy products (1-2,7). This may be detrimental in certain patients with heart or musculoskeletal diseases (1-2,7). The main adverse clinical effects of statins are at a muscular level and include increase of serum creatine kinase, sometimes associated with myalgia, rhabdomyolysis, and rare cases of inflammatory myopathy (3,8-10).
Among the marketed statins are lipid-soluble HMG-CoA reductase inhibitors that can enter the myocardial cell and thus decrease the generation of ubiquinone in these specific cells (11,12). Ichihara et al. (13) reported a specific deleterious effect of the lipid-soluble simvastatin on myocardial function in an experimental dog model. Conversely, it was reported that ubiquinone supplementation or parenteral administration could protect the myocardium or improve its altered function in certain clinical or experimental conditions (14-16). Certain studies, however, were negative (17,18).
In patients who have experienced an acute myocardial infarction, short- and long-term prognosis are strongly related to the extent of left ventricle (LV) dysfunction (19-21). Despite reassuring data from large-scale clinical trials with statins in secondary prevention (22,23), the question of a possible deleterious effect of statins on LV function in patients with established coronary heart disease (CHD) and mild to moderate LV wall-motion abnormalities, as it is seen after a Q-wave myocardial infarction, remains open.
This double-blinded randomized study was designed to assess whether reducing blood cholesterol with simvastatin in hypercholesterolemic survivors of a previous Q-wave acute myocardial infarction may affect global and segmental LV function and also exercise tolerance, which is an essential component of quality of life after an infarction. In the absence of previous data in the literature, sample size was calculated, quite arbitrarily, to test the hypothesis of a difference between groups of 5% in the rest-to-peak exercise change between before and after treatment with a power of 90% (two-sided test) and a 5% type I error. To prevent possible confounding due to the aggravation of the coronary atherosclerotic lesions during follow-up, which may by itself influence LV function and exercise testing, the study was limited to 12 weeks and only clinically stable patients were included and randomized. For ethical reasons in this population with both hypercholesterolemia and established CHD, patients randomized in the control group did not receive a placebo but a well-tolerated fibrate derivative (micronized fenofibrate) whose safety is well established (24) and with an efficacy globally similar to that of simvastatin in patients with primary hyperlipidemia (25).
MATERIALS AND METHODS
Patients with primary hypercholesterolemia (total cholesterol, >6.5 mM) and previous (>6 months) confirmed Q-wave acute myocardial infarction were enrolled. The study was approved by the Ethics Committee of the University Hospital of the Lyon city, and each patient read and signed an informed-consent form. Main exclusion criteria were signs of hepatic or renal failure, any muscular symptoms, age >70 years, female gender, refusal to follow a low-fat prudent diet or to interrupt a previously prescribed lipid-reducing treatment, blood triglycerides >3.5 mmol/L, inability to perform a symptom-limited submaximal exercise test, and clinical instability. Most of these patients were treated with cardiovascular drugs (aspirin or warfarin, β-blockers, calcium antagonists, etc.). With the agreement of the attending physicians of each patient, these treatments were not changed during the study unless necessary. In such a case, the patient was withdrawn from the study.
The study used a double-blinded design. After the washout period (8 weeks), patients were randomly assigned to either a daily 20-mg simvastatin treatment or a daily 200-mg micronized fenofibrate treatment. During the study, visits were scheduled before (selection visit) and after (inclusion visit) the washout period and after 12 weeks of treatment. Exercise testing and venticulography were performed at the inclusion visit and 12 weeks later. Physical examination, electrocardiogram, and evaluation for adverse clinical and laboratory effects were carried out at each visit. Compliance with therapy was assessed by capsule counting.
Blood samples were taken after a 12-h overnight fasting at each visit. Plasma cholesterol and triglycerides, high-density lipoprotein (HDL) cholesterol, lipoprotein (a), and safety measures (including blood cell count, uric acid, aminotransferases, glucose, creatinine, alkaline phosphatase, γ-glutamyl transferase, bilirubin, and creatine kinase) were measured at each time point. Plasma fibrinogen and ubiquinone (11) were assayed at baseline and after 12 weeks.
Exercise radionuclide ventriculography
Gated equilibrium radionuclide ventriculography was performed at rest and during bicycle exercise with the patient in the upright position, by using red blood cells labeled in vivo with ∼750 MBq of technetium-99m (activity adjusted to patient weight). High temporal resolution time-activity curves were generated, from which functional variables (times and velocities) for LV ejection and filling parameters were computed as described (26). Patients performed graded exercises that began with a workload of 25 watts and increased by 25 watts every 3 min. Imaging was performed in the last 2 min of each exercise stage and immediately after exercise. Blood pressure, heart rate, and ST segments were continuously monitored during and after exercise. Assessment of global and regional myocardial function was performed independently by two observers blinded to the identity and clinical status of the patient and to the temporal sequence in which the imaging studies were carried out. Imaging was performed in the optimal left anterior oblique view (∼45° with the best separation of the left and right cavities). End-diastolic, end-systolic, and left paraventricular background regions of interest were automatically designated and sometimes modified by the observers. End-diastolic and end-systolic volumes were deduced from the background-corrected counts at each period, and ejection fraction was calculated as the difference between the two volumes divided by the end-diastolic volume. For segmental wall-motion analysis, the left ventricle was divided into nine segments, and each segment was automatically scored according to a 3-point evaluation system in which 1 represents normokinesis; 2, hypo- or akinesis; and 3, hyperkinesis. For the global ejection fraction, the mean of the values given by the two observers was used. With linear regression analysis, a close correlation was found between the global ejection fraction (r = 0.988; mean difference, 0.32%) and end-diastolic volume (r = 0.999; mean difference, 0.11%) evaluations given by the two observers indicating that interobserver variation was very low.
Homogeneity of all continuous variables between treatment groups at baseline was analyzed by using one-way analysis of variance (ANOVA). Categoric variables were analyzed by means of χ2 testing or Fisher's Exact test. At baseline and after 12 weeks, the differences between the ejection fraction at rest and the ejection fraction at maximal exercise were calculated, and the comparison of these differences was considered the primary end point in determining a difference between groups. One-way ANOVA was used to test the null hypothesis of no difference between the two treatments. Secondary end points included the evolution of the global left ventricle ejection fraction between baseline and after 12 weeks in each group at rest and at submaximal exercise. Similar analyses were performed for the end-diastolic volume and its variations in each group between rest and peak exercise and between baseline and after 12 weeks of treatment. Any change in exercise tolerance (in watts) during follow-up was evaluated in the two groups as a secondary end point.
Biologic parameters, including those used to evaluate the safety of the treatments, were analyzed by means of similar tests (one-way ANOVA). Tolerability and adverse effects in each group were analyzed as categoric variables and compared between groups by means of χ2 testing.
Of the 64 patients randomized to treatment, 62 (30 simvastatin and 32 fenofibrate) completed the study. The mean age of the population was 54.1 years; the mean body mass index was 27.6 kg/m2; 27% of the patients were current smokers; and 31% had no physical activity (no difference between groups). In average, 67% of the patients had type IIA dyslipidemia. All randomized patients had localized (two or more hypokinetic adjacent segments) ventricular dysfunction. The two groups were homogeneous for the main clinical and biologic characteristics at baseline.
Results of the primary efficacy criterion, the comparison between groups of the change in global LV ejection fraction between rest and peak exercise measured before and after treatment, are presented in Table 1. None of the differences between groups were statistically significant. There was, however, a nonsignificant tendency toward a lower ejection fraction at rest in the simvastatin group before treatment (49.3 vs. 52.1% in the fenofibrate group), which did not persist after 12 weeks (52.1 vs. 52.9%). The difference between the values measured at rest before and after 12 weeks in the simvastatin group was significant (p = 0.009). To assess whether simvastatin may have been beneficial specifically in patients with LV dysfunction, subgroup analyses were performed after separating patients with ejection fraction lower or higher than 52.5%, the median value in the two groups. We also analyzed the effect of simvastatin in patients with ejection fraction lower or higher than 40%, a value often considered to reflect severe LV dysfunction. In the subgroup with ejection fraction <52.5% (n = 15), ejection fraction increased from 41.5 ± 8.2% before simvastatin to 46.3 ± 10% after simvastatin, whereas in the subgroup with ejection fraction >52.5%, ejection fraction did not change after simvastatin (from 60.2 ± 5.6% to 60.9 ± 5.8%). The corresponding changes in the fenofibrate subgroups were from 42.8 ± 8.5% to 45.0 ± 9.5% and from 61.4 ± 7.1% to 60.8 ± 7.2%. In the subgroups with ejection fraction <40% (n = 8), ejection fraction increased from 33.2 ± 5.2% to 39.4 ± 12.6% with simvastatin against from 54.2 ± 8.6% to 56.0 ± 7.7% in the subgroup with ejection fraction >40%. In the fenofibrate subgroups, the changes were from 35.2 ± 6.7% to 36.6 ± 8.1% and from 56.9 ± 8.6% to 57.5 ± 7.5%. Thus it seems that simvastatin was beneficial essentially in patients with marked LV dysfunction. As the sample size was small (it was not calculated to make subgroup analyses), statistic calculations were not performed.
LV segmental function in the two groups was then analyzed to examine the relative number of segments that were normo-, hypo-, or hyperkinetic before and after treatment, at rest, and at peak exercise, in the two groups. We did not perform subgroup analyses. In the fenofibrate group, there were 130 hypokinetic segments at rest before treatment, 140 after 12 weeks, and 130 and 118 at peak exercise, respectively. The values in the simvastatin group were 128 and 132 segments at rest and 100 and 91 segments at peak exercise. There was no significant difference between groups. Regarding the number of normokinetic and hyperkinetic segments and their evolution after treatment, there also was no difference between the two treatment groups. Table 2 presents results of a secondary end point, the LV end-diastolic volume. Data show similar trends and no statistically significant differences except for the comparison between values at rest before and after 12 weeks of simvastatin (p = 0.01).
Results concerning exercise testing are shown in Table 3 and indicate no difference between groups at baseline and no significant effect of both treatments on the tolerance to exercise. This is confirmed by the observed changes in heart rate and blood pressure before and after treatment (data not shown).
The main biologic effects of the two treatments are presented in Table 4. We observed significant reductions of total and LDL cholesterol in both groups, no effect on HDL cholesterol, significant reduction of triglycerides and fibrinogen only in the fenofibrate group, and significant reduction of serum ubiquinone only in the simvastatin group. When ubiquinone was expressed as a ratio of ubiquinone to LDL cholesterol, the difference between groups did not remain significant (p = 0.13).
Regarding the tolerability of both drugs and the occurrence of adverse clinical events, both drugs were generally well tolerated. In particular, no patient reported any adverse musculoskeletal symptoms. Three fenofibrate and one simvastatin patients reported transient dyspepsia, which did not result in interruption of the treatment. Two patients in the simvastatin group did not complete the study because of coronary events not imputable to treatment and which required hospitalization and interruption of treatment. Finally, during follow-up, no abnormal biologic value (value >3 times the upper limit of normal for liver enzymes and creatine kinase) was recorded.
Summary of the data
The results of this study indicate that a 12-week treatment with either 20 mg simvastatin or 200 mg micronized fenofibrate did not result in any further alteration in global LV function at rest and during exercise in dyslipidemic patients with established CHD.
Regarding the biologic effects of the treatment, only simvastatin reduced serum ubiquinone levels, although the difference between groups did not remain significant when ubiquinone level was expressed as a ratio to the LDL cholesterol level. This is not surprising because serum ubiquinone is thought to be essentially transported by LDL, and simvastatin treatment resulted in a slightly larger decrease in LDL than did fenofibrate. On the other hand, it is noteworthy that fenofibrate treatment did not result in any decrease in ubiquinone despite significant decrease in LDL, which does not seem to be in agreement with the LDL-ubiquinone theory. The data even suggest that the amount of ubiquinone per LDL molecule was increased in the fenofibrate group, which could be in agreement with the report that fibrate derivatives increase the synthesis of ubiquinone (27). Further studies are thus warranted to examine more specifically the relations between lipoproteins and ubiquinone, as well as their alterations by the lipid-reducing interventions. It is noteworthy, however, that the ratio of ubiquinone to LDL has been presented as an important risk factor of atherosclerosis (28). Ubiquinone is indeed considered a potent antioxidant (29) that may protect LDL from oxidative alteration (30). It can therefore be said that, in terms of prevention of atherosclerosis through alteration of this ratio, the two treatments tested in that study may be complementary. Actually, if the statin reduces LDL more than the fibrate but alters ubiquinone, fenofibrate seems to increase ubiquinone. In turn, fenofibrate, but not simvastatin, reduces triglycerides and fibrinogen, two major risk factors for CHD. A logical conclusion should be that the association of the two agents may constitute an original and potent cardioprotective approach, especially in coronary patients with mixed dyslipidemia. An important condition, however, should be that for a given dosage of each drug, no major deleterious side effects would occur, which remains to be demonstrated. Further studies are warranted to examine the point.
Simvastatin and left ventricular dysfunction
Whereas the long-term (>12 weeks) effect of cholesterol-reducing drugs on serum and tissue ubiquinone levels as well as on myocardial function should be examined in future studies, our data suggest that simvastatin (as well as fenofibrate) treatment had no obvious deleterious effects on the hearts of patients with established CHD. In the simvastatin group, there even was a significant improvement in the rest LV ejection fraction and end-diastolic volume, suggesting an unexpected beneficial effect of simvastatin on heart function. However, as the rest LV ejection fraction was not the primary end point of the study (sample size was not calculated for that purpose), that result may be due to chance and should be confirmed in future studies. On the other hand, although this effect of simvastatin was not expected, it was highly statistically significant. This beneficial effect may be clinically relevant, as recent evidence from the 4S study indicated that the incidence of heart failure was decreased with simvastatin (31). Finally, is there any biologic rationale to explain that effect? Several clinical studies recently showed that statin treatment is associated with a restoration of endothelial function (32,33) and improvement of myocardial perfusion (34,35), as well as myocardial ischemia (36-38). In this view, the contribution of the coronary microcirculation and the effect of lipid reduction should be considered. Indeed, Gould et al. (35), who used positron emission tomography, demonstrated a decrease in the size and the severity of myocardial-perfusion abnormalities after short-term cholesterol reduction. Interestingly, in our study, the patients with marked LV dysfunction (ejection fraction <40%), those patients with theoretically the more severe atherosclerosis and the more altered endothelial function, benefited the most from the statin treatment, although that point must be confirmed in studies specifically designed for that purpose.
Another important point from this study is that for a reduction of blood total and LDL cholesterol close to that of simvastatin, fenofibrate did not improve LV function at rest as simvastatin did. This may suggest (but did not prove) that it was not the reduction of LDL cholesterol per se that was beneficial but that the increased ejection fraction resulted from a specific effect of simvastatin independent of the reduction of LDL. This may be (if it is confirmed) a major point, quite in contradiction to the current opinion (39) about the importance of reducing LDL to improve the prognosis in secondary prevention of CHD. It also emphasizes the idea that "statins do more than just lower cholesterol" (40,41). Our study is nonetheless the first to suggest that statin treatment may improve ventricular function.
Finally, it is noteworthy that the effect of larger dosages also should be tested in future trials, as many patients are treated with simvastatin dosage >20 mg per day. It should also be noted that the improvement in LV ejection fraction at rest after simvastatin was associated with a lack of increase of ejection fraction during exercise, whereas before treatment, there was a significant increase at peak exercise in that group, as in the fenofibrate group before and after treatment. This observation should not be neglected and may suggest that simvastatin had mobilized at rest a myocardial reserve (an ability to vasodilate). This adaptive mechanism likely occurs only during exercise in the absence of treatment. If this effect of simvastatin is confirmed, its mechanism should be carefully investigated, as it may open a new avenue for the management of various heart-failure syndromes associated with CHD or dyslipidemias or both.
Limitations of the study
A possible limitation of the study is its relative brevity, as it could be argued that 12 weeks may be too short a period to expect significant changes in myocardial function. On the other hand, O'Driscoll et al. (33) reported a significant improvement of endothelial function within a 1-month simvastatin treatment, which is in line with the effect of simvastatin on the LV rest ejection fraction in our study. Thus it seems that a 5- to 12-week treatment was long enough to detect rather small changes in LV function in these patients. Similar reasoning could apply to the question of a possible lack of sensitivity of radionuclide ventriculography to detect small changes in LV function.
All the patients randomized in this study were survivors of a well-defined Q-wave acute myocardial infarction, and LV segmental dysfunction was observed on the ventriculography of all patients before treatment. Thus the lack of evident deleterious effect on the left ventricle of the patients treated with simvastatin (and also with micronized fenofibrate) in this study suggests that both drugs are safe for the postinfarct myocardium, at least at those dosages. These data even suggest that simvastatin may improve LV function, in particular, in patients with marked LV dysfunction. Although these data must be confirmed, they support the view that certain cholesterol-reducing interventions (at least, those that are able to influence endothelial function) are indicated in patients with altered LV function, as was suggested in recent successful clinical trials with statins (22,23). Further studies are warranted to solve this question definitively.
Acknowledgment: The study was supported by a grant from the Laboratoire Fournier, Daix, France.
1. De Pigneux G, Chariot P, Ammi-Saïd M, et al. Lipid-lowering drugs and mitochondrial function: effects of HMGCoA reductase inhibitors on serum ubiquinone and blood lactate/pyruvate ratio. J Clin Pharmacol
2. Folkers K, Langsjoen P, Willis R, et al. Lovastatin decreases coenzyme Q levels in humans. Proc Natl Acad Sci U S A
3. London SF, Gross KF, Ringel SP. Cholesterol-lowering agent myopathy. Neurology
4. Faust J, Goldstein J, Brown M. Synthesis of ubiquinone and cholesterol in human fibroblasts: regulation of a branched pathway. Arch Biochem Biophys
5. Grundy S. HMGCoA reductase inhibitors for treatment of hypercholesterolemia. N Engl J Med
6. Goldstein JL, Brown MS. Regulation of mevalonate pathway. Nature
7. Langsjoen PH, Vadhanavikit S, Folkers K. Response of patients in classes III and IV of cardiomyopathy to therapy in a blind and crossover trial with coenzyme Q10. Proc Natl Acad Sci U S A
8. Stalenhoef AF, Mol MJ, Stuyt PM. Efficacy and tolerability of simvastatin. Am J Med
9. Deslypere JP, Vermeulen A. Rhabdomyolysis and simvastatin. Ann Intern Med
10. Schalke BB, Schmidt B, Toyka K, Hartung HP. Pravastatin-associated inflammatory myopathy. N Engl J Med
11. Belichard P, Pruneau D, Zhiri A. Effect of long-term treatment with lovastatin or fenofibrate on hepatic and cardiac ubiquinone levels in cardiomyopathic hamster. Biochim Biophys Acta
12. Satoh K, Yamato A, Nakai T, Hoshi K, Ichihara K. Effects of 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors on mitochondrial respiration in ischaemic dog hearts. Br J Pharmacol
13. Ichihara K, Satoh K, Abiko Y. Influences of pravastatin and simvastatin, HMGCoA reductase inhibitors, on myocardial stunning in dogs. J Cardiovasc Pharmacol
14. Takeo S, Tanonaka K, Tazuma Y, Miyake K, Murai R. Possible mechanism by which coenzyme Q10 improves reoxygenation-induced recovery of cardiac contractile force after hypoxia. J Pharmacol Exp Ther
15. Matsushima T, Sueda T, Matsuura Y, Kawasaki T. Protection by coenzyme Q10 of canine myocardial reperfusion injury after preservation. J Thorac Cardiovasc Surg
16. Greenberg S, Frishman WH. Coenzyme Q10: a new drug for cardiovascular disease. J Clin Pharmacol
17. Birnbaum Y, Hale SL, Kloner RA. The effect of coenzyme Q10 on infarct size in a rabbit model of ischemia/reperfusion. Cardiovasc Res
18. Permanetter B, Rössy W, Klein G, Weingartner F, Seidl KF, Blömer H. Ubiquinone (coenzyme Q10) in the long-term treatment of idiopathic dilated cardiomyopathy. Eur Heart J
19. Multicenter Postinfarction Research Group. Risk stratification and survival after myocardial infarction. N Engl J Med
20. Lee KL, Pryor DB, Pieper KS, et al. Prognostic value of radionuclide angiography in medically treated patients with coronary artery disease: a comparison with clinical and catheterization variables. Circulation
21. Loh E, Sutton M, Wun C-C, et al. Ventricular dysfunction and the risk of stroke after myocardial infarction. N Engl J Med
22. Scandinavian Simvastatin Survival Study group. Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4S). Lancet
23. Sacks FM, Pfeffer MA, Moye LA, et al. for the Cholesterol and Recurrent Events Trial investigators, the effect of pravastatin on coronary events after myocardial infarction in patients with average cholesterol levels. N Engl J Med
24. Rössner S, Orö L. Fenofibrate therapy of hyperlipoproteinaemia. Atherosclerosis
25. Farnier M, Bonnefous F, Debbas N, Irvine A. Comparative efficacy and safety of micronized fenofibrate and simvastatin in patients with primary type IIa or IIb hyperlipidemia. Arch Intern Med
26. Itti R, Bontemps L. Explorations isotopiques cardiaques. In: Encyclopédie médicochirurgicale: cardiologie-angiologie.
Paris: El-sevier, 1996:8.
27. Aberg F, Zhang Y, Appelkvist EL, Dallner G. Effects of clofibrate, phthalates and probucol on ubiquinone levels. Chem Biol Interact
28. Hanaki Y, Sugiyama S, Ozawa T, Ohno M. Ratio of low-density lipoprotein cholesterol to ubiquinone as a coronary risk factor. N Engl J Med
29. Beyer RE. The participation of coenzyme Q in free radical production and antioxidation. Free Radic Biol Med
30. Stocker R, Bowry VW, Frei B. Ubiquinol-10 protects human low density lipoprotein more efficiently against lipid peroxidation than does alpha-tocopherol. Proc Natl Acad Sci U S A
31. Van Boven AJ, Brugemann J, De Graef PA, May JF, Crijns HJGM. The 4S study: implications for prescribing. Drugs
32. Treasure CB, Klein JL, Weintraub WS, et al. Beneficial effect of cholesterol-lowering therapy on the coronary endothelium in patients with coronary artery disease. N Engl J Med
33. O'Driscoll G, Bao BCH, Green D, Taylor RR. Simvastatin, an HMG-CoA reductase inhibitor, improves endothelial function within 1 month. Circulation
34. Eichstadt HW, Eskotter H, Hoffmann I, Amthauer HW, Weidinger G. Improvement of myocardial perfusion by short-term fluvastatin treatment in coronary artery disease. Am J Cardiol
35. Gould KL, Martucci JP, Goldberg DI, et al. Short term cholesterol lowering decreases size and severity of perfusion abnormalities by positron emission tomography after dipyridamole in patients with coronary artery disease. Circulation
36. Andrews TC, Raby K, Barry J, et al. Effect of cholesterol reduction on myocardial ischemia in patients with coronary disease. Circulation
37. Van Boven AJ, Jukema JW, Zwinderman AH, Crijns HJGM, Lie KI, Bruschke AVG. Reduction of transient myocardial ischemia with pravastatin in addition to the conventional treatment in patients with angina pectoris. Circulation
38. De Divitiis M, Rubba P, Di Somma S, et al. Effect of short term reduction in serum cholesterol with simvastatin in patients with stable angina pectoris and mild to moderate hypercholesterolemia. Am J Cardiol
39. Frye RL. Clinical reality of lowering total and LDL cholesterol. Circulation
40. Vaughan CJ, Murphy MB, Buckley BM. Statins do more than just lower cholesterol. Lancet
41. Rosenson RS, Tangney CC. Antiatherothombotic properties of statins: implications for cardiovascular event reduction. JAMA