Coenzyme Q₁₀ (ubiquinone) is a naturally occurring, lipid-soluble, quinone which, by acting as an electron transporter, is an essential cofactor in mitochondrial oxidative phosphorylation and generation of adenosine triphosphate (1- 2). In its reduced form, coenzyme Q₁₀ is also thought to act as a lipophilic antioxidant protecting cell membranes and lipoproteins in the circulation from oxidation (1- 3). About one-half of coenzyme Q₁₀ is ingested in dietary fat, and the remainder is synthesized endogenously through the mevalonate pathway, which is blocked by statins (1- 3).

Coenzyme Q₁₀ deficiency has been associated with myopathy, and there has been concern that statins might cause peripheral and cardiac muscle dysfunction by reducing coenzyme Q₁₀ production (4- 5). In theory, coenzyme Q₁₀ depletion could lead to muscle energy starvation (a particular concern in the failing heart [6]) and oxidative damage to myocytes. These theoretical concerns have been coupled with the observation that low cholesterol is associated with a worse prognosis in heart failure (7), forming the basis of articles in the lay press and on the web that have suggested that statins might be dangerous in heart failure. In practice, however, the role of coenzyme Q₁₀ in the effect of statins on muscle function (if any) is uncertain, as is the association between plasma coenzyme Q₁₀ concentration and clinical outcomes in cardiovascular disease (8- 18). In 1 recent study, however, low plasma coenzyme Q₁₀ concentration was found to be an independent predictor of mortality in patients hospitalized with heart failure (18).

Because of the concerns alluded to above, the U.S. Food and Drug Administration requested that we measure plasma coenzyme Q₁₀ concentration in a subset of the patients with ischemic systolic heart failure enrolled in the CORONA (Controlled Rosuvastatin Multinational Study in Heart Failure) trial. In this pre-specified substudy, we investigated the effect of statin therapy on coenzyme Q₁₀ concentration, as well as the relationship between coenzyme Q₁₀ and fatal and nonfatal cardiovascular events (19- 20).

The design and principal findings of the CORONA study have been reported in detail (19- 20). Patients ≥60 years of age with chronic New York Heart Association (NYHA) functional class II to IV heart failure of investigator-reported ischemic etiology and a left ventricular ejection fraction (LVEF) of ≤40% (≤0.35 if NYHA functional class II) were eligible, provided that the investigator believed they did not need treatment with a cholesterol-lowering drug.

Exclusion criteria included recent cardiovascular events, procedures, or operations (or planned procedures or operations); acute or chronic liver disease or alanine aminotransferase >2 times the upper limit of normal (ULN); serum creatinine >220 μmol/l (>2.49 mg/dl); chronic muscle disease or unexplained creatine kinase >2.5 times ULN; thyroid-stimulating hormone >2 times ULN; or any other condition substantially reducing life expectancy.

The trial was approved by ethics committees of participating hospitals, and patients provided written informed consent. Patients were allocated, equally, to 10 mg of rosuvastatin or matching placebo, once daily. We measured serum creatinine, creatine kinase, thyroid-stimulating hormone, alanine aminotransferase, high-sensitivity C-reactive protein, and lipid/lipoproteins (total, low-density lipoprotein [LDL] cholesterol, high-density lipoprotein cholesterol, triglycerides, and apolipoprotein [apo] A-1 and B) at baseline in all 5,011 patients.

After the study started, the protocol was amended to include measurement of N-terminal pro–B-type natriuretic peptide (NT-proBNP), which was available in 3,664 (73%) patients. All measurements, except thyroid-stimulating hormone, were repeated at 3 months. In a pre-specified substudy, coenzyme Q₁₀ was measured in 1,191 patients using a high-performance liquid chromatography method after extraction of serum into hexane and using vitamin K₁ as an internal standard. The reference range is 0.34 to 2.54 μg/ml (0.39 to 2.94 μmol/l).

All blood samples were nonfasting and were analyzed at a central laboratory (Medical Research Laboratories, Zaventem, Belgium). The LDL was directly measured. Coenzyme Q₁₀ was analyzed on fresh samples sent at refrigerated temperature by overnight mail to the central laboratory.

The primary outcome was the composite of cardiovascular mortality, nonfatal myocardial infarction, or nonfatal stroke, analyzed as time to the first event. The secondary outcomes were (in listed order): all-cause mortality, any coronary event (defined as sudden death, fatal or nonfatal myocardial infarction, percutaneous coronary intervention, coronary artery bypass graft surgery, ventricular defibrillation by an implantable cardioverter-defibrillator, resuscitation from cardiac arrest, or hospitalization for unstable angina), cardiovascular mortality (cause-specific cardiovascular death was also analyzed), and number (episodes) of hospitalizations (for cardiovascular causes, unstable angina, and worsening heart failure). The present report focuses on the primary end point, total mortality, the coronary end point, and hospitalizations (all-cause, cardiovascular cause, and worsening heart failure). We also included the additional post-hoc composite outcome of death from any cause or hospitalization for worsening heart failure (analyzed as time to first event) because of previously expressed concerns that coenzyme Q₁₀ deficiency might cause worsening heart failure, leading to increased risk of hospital admission and death. We conducted further post-hoc analyses of patients hospitalized for all causes, cardiovascular causes, worsening heart failure, and noncardiovascular causes (analyzed as time to first event). The definition and adjudication of all outcomes have been described in detail previously (19- 20). As the result of a protocol amendment adopted 15 months after the start of the trial, patients also completed a questionnaire about muscle symptoms at each study visit and had a measurement of creatine kinase at 6 and 15 months after randomization, yearly thereafter, and at the last study visit (20). Patients were asked 2 questions: whether they had any muscular pain since the previous visit, and whether they had muscular pain at the present visit.

We addressed 2 main questions: 1) Was baseline serum coenzyme Q₁₀ concentration associated with the range of clinical outcomes described above? 2) Did treatment with rosuvastatin increase the risk of any of the described outcomes in patients with a low serum coenzyme Q₁₀ concentration?

To answer the first question, we examined clinical characteristics and outcomes in patients divided according to tertile of baseline coenzyme Q₁₀ concentration, and we entered baseline coenzyme Q₁₀ (as a continuous variable) in a series of extensive multivariable models previously developed in the CORONA study population (21). These models had been built for the other mortality–morbidity composite outcomes listed in the previous text, in addition to all-cause mortality.

To answer the second question, we examined the effect of rosuvastatin compared with placebo in each of the baseline coenzyme Q₁₀ tertiles, looking at both the unadjusted treatment effect and effect of treatment adjusted for age group (≥75/<75 years); sex (female/male); baseline LVEF (≥0.25/<0.25) and NYHA functional class (III to IV/II); beta-blocker use (yes/no); total cholesterol (≥6.0/<6.0 mmol/l); and history of myocardial infarction (yes/no) or hypertension (yes/no), as pre-specified in the main CORONA study analysis plan. Tests for interaction between treatment effect and baseline coenzyme Q₁₀ value were carried out as described in the following text.

For continuous variables, differences in baseline variables between the patients in each coenzyme Q₁₀ tertile were tested with the Student t test (NT-proBNP with the Wilcoxon rank-sum test) and for categorical variables with the Fisher exact test. For comparison of tertiles, we used the Jonckheere-Terpstra test (for continuous variables) and the Cochran-Armitage trend test (for categorical variables). The multivariable analyses to which baseline coenzyme Q₁₀ concentration was added as a continuous variable have been described in detail previously (21). The 8 most important demographic and clinical variables included age, sex, LVEF, NYHA functional class, heart rate, body mass index, history of diabetes mellitus, and intermittent claudication. The 2 most important biochemical variables were serum creatinine concentration and apoA-1 concentration. The log concentration of the neurohumoral marker NT-proBNP was the single most important predictor of all outcomes (21).

Cox's proportional hazards models (unadjusted and adjusted) were used to calculate hazard ratios and 95% confidence intervals (SAS version 8.2, SAS Institute, Cary, North Carolina) in all patients and in each coenzyme Q₁₀ concentration tertile separately. The adjusted Cox regression model incorporated randomized treatment and the variables described earlier. Similar Cox analyses were performed to compare cardiovascular risk between NT-proBNP tertile 1 and tertile 3 in the placebo group.

Total number (episodes) of hospital admissions were analyzed using a permutation test. Tests for interaction between treatment effect and coenzyme Q₁₀ tertile, for each outcome, were carried out using a Cox regression analysis with the following covariates, treatment as 0/1, coenzyme Q₁₀ tertile as 0/1, and treatment*coenzyme Q₁₀ tertile (interaction) as 0/1. We also analyzed interaction by treatment with coenzyme Q₁₀ as a continuous variable.

The baseline characteristics of the 1,191 patients with a measurement of coenzyme Q₁₀ are shown in (Table 1) (all patients and by tertiles of baseline coenzyme Q₁₀ concentration).

Patients in the lowest coenzyme Q₁₀ concentration tertile (tertile 1) were, on average, older, in a higher NYHA functional class, had more atrial fibrillation/flutter, had lower plasma lipids, and had a lower LVEF and estimated glomerular filtration rate compared with those in the highest tertile. NT-proBNP concentration was also significantly higher in patients in the lowest coenzyme Q₁₀ tertile compared with the highest tertile.

In the whole group of patients studied, LDL declined from a mean of 142 mg/dl at baseline to 76 mg/dl at 3 months with rosuvastatin but did not change in the placebo group: 141 mg/dl at baseline and 141 mg/dl at 3 months (48% net difference; p < 0.0001). The corresponding net difference in tertiles of coenzyme Q₁₀ was 51%, 48%, and 45% (tertiles 1, 2, and 3, respectively).

Overall, coenzyme Q₁₀ also declined at 3 months with rosuvastatin but did not change in the placebo group (39% net difference; p < 0.0001) (Table 2). Rosuvastatin reduced plasma coenzyme Q₁₀ concentration in all 3 tertiles (Table 2).

In patients treated with placebo, the risk of the pre-defined primary outcome of cardiovascular death, myocardial infarction, or stroke (expressed as patients experiencing an event per 100 person-years of follow-up) was numerically highest in patients in the lowest coenzyme Q₁₀ tertile, intermediate in the middle tertile, and lowest in patients in the highest coenzyme Q₁₀ tertile (Table 3). The same relationship was seen between coenzyme Q₁₀ tertile and mortality, the other composite outcomes, and hospitalizations (Table 4). However, risk was not significantly higher in coenzyme Q₁₀ tertile 1, compared with tertile 3, after adjustment for other prognostic variables (Table 5).

When entered as a continuous variable in our previously described multivariable models, coenzyme Q₁₀ was not an independent predictor of all-cause mortality (Table 6), or any of the other mortality–morbidity outcomes examined (data not shown).

The hazard ratio estimating the treatment effect for all 4 time-to-first-event end points was >1.0 in coenzyme Q₁₀ tertile 1 and <1.0 in tertiles 2 and 3, although the 95% confidence intervals overlapped 1.0 in all subgroups (Table 3). There was no significant interaction between treatment effect and coenzyme Q₁₀ tertile, with p values ranging from 0.14 (primary end point) to 0.26 (coronary end point) for the 4 end points (Table 3). Corresponding p values with coenzyme Q₁₀ included as a continuous variable were between 0.49 (coronary end point) and 0.95 (primary end point).

A similar picture was seen when the post-hoc outcome of number of patients hospitalized (analyzed as time to first hospitalization) was examined, with rosuvastatin:placebo hazard ratios >1 in coenzyme Q₁₀ tertile 1 for hospitalizations for any cause, cardiovascular causes, worsening heart failure, and noncardiovascular causes (Table 4). All corresponding hazard ratios in tertiles 2 and 3 were <1. As with all-cause mortality and the composite mortality–morbidity outcomes, there was no statistically significant interaction between treatment and coenzyme Q₁₀ tertile, although the p value for cardiovascular hospitalization was 0.052.

By contrast, the total number (episodes) of hospitalizations (a pre-specified secondary end point) for admissions due to any cause, cardiovascular causes, and heart failure were similar in the 2 treatment groups in tertile 1 and generally numerically fewer in the rosuvastatin group in tertiles 2 and 3 (Table 4).

Although there were no statistically significant differences between the treatment groups in the coenzyme Q₁₀ tertile 1, there was an excess of 11 deaths in the rosuvastatin group compared with placebo group (Table 3). There were 5 extra cardiovascular deaths, 3 of which were due to myocardial infarction and 2 of which were sudden. There were 6 extra noncardiovascular deaths. The number of deaths due to worsening heart failure was 21 in each treatment group in coenzyme Q₁₀ tertile 1.

Looking at nonfatal events, there were 18 more patients hospitalized at least once for any reason in the rosuvastatin group compared with placebo group; however, only 3 more episodes of hospital admissions. The equivalent numbers for cardiovascular hospitalization were +13 and −5. One fewer patient in the rosuvastatin group than the placebo group was hospitalized for worsening heart failure (and there were 8 fewer admissions for heart failure in the rosuvastatin group). There were 9 more nonfatal myocardial infarctions in the rosuvastatin group in coenzyme Q₁₀ tertile 1.

The mean change in NYHA functional class from baseline to last study visit in coenzyme Q₁₀ tertile 1 was −0.085 in the placebo group and −0.035 in the rosuvastatin group (p = 0.44). The equivalent changes in tertile 2 were −0.015 and −0.104 (p = 0.14), and in tertile 3, they were −0.122 and −0.145 (p = 0.75).

Similar numbers of patients in each coenzyme Q₁₀ tertile reported muscular pain on questioning, and this was also the case for placebo-treated compared with rosuvastatin-treated patients. For the question about muscular pain since the previous visit, the proportions in the placebo group were 7.5%, 10.3%, and 8.1% (tertiles 1, 2, and 3, respectively), and in the rosuvastatin group, they were 8.5%, 9.8%, and 6.3%, respectively. Corresponding figures for the question about muscular pain at the current visit, were 6.0%, 7.4%, and 6.6% (placebo) and 6.0%, 6.0%, and 4.3% (rosuvastatin). Only 1 patient had a creatine kinase value >10 times ULN during follow-up (randomized to placebo).

The number of patients in coenzyme Q₁₀ tertile 1 who discontinued study drug for any reason was 50 (30 because of an adverse event) in the placebo group and 44 (24 because of an adverse event) in the rosuvastatin group. The equivalent numbers in tertile 2 were 54 (35 because of an adverse event) and 38 (19 because of an adverse event), and in tertile 3, the numbers were 52 (31 because of an adverse event) and 44 (28 because of an adverse event).

We found that patients with a lower serum coenzyme Q₁₀ concentration at baseline were older and had evidence of more severe heart failure. In particular, several powerful predictors of poor prognosis were more prevalent in patients with a lower coenzyme Q₁₀, including lower LVEF and estimated glomerular filtration rate and higher NYHA functional class and NT-proBNP concentration. Lower coenzyme Q₁₀ was also associated with higher age and, as expected, lower lipid levels (lower lipid levels are also a marker of poor prognosis in heart failure). Although lower coenzyme Q₁₀ was associated with a higher risk of death in unadjusted analyses, coenzyme Q₁₀ concentration was not an independent predictor of mortality in a multivariable analysis (or an independent predictor of any other outcome). This finding differs from that of the 1 other study investigating the relationship between coenzyme Q₁₀ and mortality in patients with heart failure (18).

There are several important differences between that report of Molyneux et al. (18) and the present study. Our study was much larger with more patients (1,191 vs. 236) and deaths (350 vs. 76). Indeed, there were twice as many deaths in the lowest coenzyme Q₁₀ tertile in our study as in the whole cohort studied by Molyneux et al. (18). That study stored plasma for up to 5.4 years before measurement of coenzyme Q₁₀. Concentration of coenzyme Q₁₀ falls with storage, and that may explain the lower levels of coenzyme Q₁₀ in the study of Molyneux et al. (12,18). Another important difference was in the multivariable analyses performed. Molyneux et al. (18) adjusted for 5 baseline variables in addition to coenzyme Q₁₀. We adjusted for 14 previously identified independent predictors of outcome (21). Coenzyme Q₁₀ may have been an independent predictor of death in the study of Molyneux et al. (18) only because they did not fully adjust for differences in other prognostic variables between patients with a lower or higher coenzyme Q₁₀ concentration. For example, when we repeated the limited Cox proportional hazards analysis described by Molyneux et al. (18) using median coenzyme Q₁₀ concentration, we found coenzyme Q₁₀ to be an independent predictor of mortality (p = 0.048; data not shown), although this was not the case after fuller adjustment (Table 6).

We also examined composites of fatal and nonfatal events, including that of death or hospital admission for heart failure, in view of prior concerns that low coenzyme Q₁₀ might lead to worsening heart failure. As with mortality, coenzyme Q₁₀ concentration was not an independent predictor of any of these other outcomes.

As expected, treatment with rosuvastatin reduced serum coenzyme Q₁₀ concentration. In view of prior concerns that statin-induced reductions in coenzyme Q₁₀ might be harmful in heart failure, we examined the effect of rosuvastatin on clinical outcomes according to baseline serum coenzyme Q₁₀ concentration. Tertile analysis showed a numerically higher event rate in statin-treated compared with placebo-treated patients for all outcomes in patients with the lowest coenzyme Q₁₀ concentration. In the other 2 coenzyme Q₁₀ tertiles, rosuvastatin treatment was associated with a numerically lower event rate than placebo treatment. None of the tests for interaction between baseline coenzyme Q₁₀ concentration tertile and treatment was statistically significant, although this is a test with low power and the p values were borderline, ranging from 0.14 to 0.26.

Although we cannot completely exclude an adverse effect of statin treatment in heart failure patients with a low coenzyme Q₁₀ concentration, we believe that several observations make such an effect unlikely. First, we could not demonstrate that low coenzyme Q₁₀ concentration, whether “spontaneous” or statin-induced, was independently associated with worse outcome in the multivariable analyses described above. Second, close inspection of outcomes in the lowest coenzyme Q₁₀ tertile did not show any evidence of increased risk of the “expected” clinical events, namely, death due to heart failure or heart failure hospitalization in the rosuvastatin group compared with the placebo group. In addition, there was no worsening of NYHA functional class in rosuvastatin-treated patients compared with placebo-treated patients in the lowest coenzyme Q₁₀ tertile. Indeed, if there was an excess of any type of event, it was myocardial infarction and noncardiovascular death. Third, we did not find any evidence of the most predicted coenzyme Q₁₀-related effect of statins, namely, muscle symptoms or increased creatine kinase. Furthermore, there were no more discontinuations of rosuvastatin than placebo in the lowest coenzyme Q₁₀ tertile.

Although a low serum coenzyme Q₁₀ concentration is associated with worse outcomes in heart failure, that is because it is a marker of more advanced disease and is not an independent predictor of prognosis. Statin treatment reduced serum coenzyme Q₁₀ concentration, but even in patients with a low starting coenzyme Q₁₀, statin therapy was not associated with a significantly worse outcome, although we had limited statistical power to completely exclude this possibility. Although we cannot completely exclude an interaction between coenzyme Q₁₀ concentration and the effect of statins, no expected or consistent pattern of harm was observed.