A Multicenter, Randomized, Placebo‐Controlled Trial of Atorvastatin for the Primary Prevention of Cardiovascular Events in Patients With Rheumatoid Arthritis

Objective Rheumatoid arthritis (RA) is associated with increased cardiovascular event (CVE) risk. The impact of statins in RA is not established. We assessed whether atorvastatin is superior to placebo for the primary prevention of CVEs in RA patients. Methods A randomized, double‐blind, placebo‐controlled trial was designed to detect a 32% CVE risk reduction based on an estimated 1.6% per annum event rate with 80% power at P < 0.05. RA patients age >50 years or with a disease duration of >10 years who did not have clinical atherosclerosis, diabetes, or myopathy received atorvastatin 40 mg daily or matching placebo. The primary end point was a composite of cardiovascular death, myocardial infarction, stroke, transient ischemic attack, or any arterial revascularization. Secondary and tertiary end points included plasma lipids and safety. Results A total of 3,002 patients (mean age 61 years; 74% female) were followed up for a median of 2.51 years (interquartile range [IQR] 1.90, 3.49 years) (7,827 patient‐years). The study was terminated early due to a lower than expected event rate (0.70% per annum). Of the 1,504 patients receiving atorvastatin, 24 (1.6%) experienced a primary end point, compared with 36 (2.4%) of the 1,498 receiving placebo (hazard ratio [HR] 0.66 [95% confidence interval (95% CI) 0.39, 1.11]; P = 0.115 and adjusted HR 0.60 [95% CI 0.32, 1.15]; P = 0.127). At trial end, patients receiving atorvastatin had a mean ± SD low‐density lipoprotein (LDL) cholesterol level 0.77 ± 0.04 mmoles/liter lower than those receiving placebo (P < 0.0001). C‐reactive protein level was also significantly lower in the atorvastatin group than the placebo group (median 2.59 mg/liter [IQR 0.94, 6.08] versus 3.60 mg/liter [IQR 1.47, 7.49]; P < 0.0001). CVE risk reduction per mmole/liter reduction in LDL cholesterol was 42% (95% CI −14%, 70%). The rates of adverse events in the atorvastatin group (n = 298 [19.8%]) and placebo group (n = 292 [19.5%]) were similar. Conclusion Atorvastatin 40 mg daily is safe and results in a significantly greater reduction of LDL cholesterol level than placebo in patients with RA. The 34% CVE risk reduction is consistent with the Cholesterol Treatment Trialists’ Collaboration meta‐analysis of statin effects in other populations.


INTRODUCTION
Despite major advances in therapy over the last two de cades, rheumatoid arthritis (RA) continues to be associated with reduced life expectancy compared to the general population (1). Almost half of all deaths in RA (~35-40% of the excess deaths) are attributed to cardiovascular disease (CVD) (2). There are many mechanisms that may underlie the increased CVD morbidity and mortality in RA, but their cross talk and relative contributions are not yet fully elucidated. CVD risk factors, including smoking, hyper tension, dyslipidemia, increased adiposity, and reduced physical activity, are highly prevalent in RA (3) but do not fully account for the excess CVD (4). A significant part is attributed to "novel" CVD risk factors, such as "high grade" inflammation promoting athero thrombotic cardiovascular events (CVEs) (4,5). Risk algorithms developed for the general population may underestimate CVE risk in patients with RA (6)(7)(8), even when multipliers are applied, as in recently updated European League Against Rheumatism rec ommendations (9). This makes identification of RA patients who would benefit from primary prevention therapy less precise, leads to significant underuse of statins even in patients who fulfill general population thresholds for statin treatment (10), and has led some to suggest universal prescription of statins in RA (11), as practiced in diabetes mellitus (DM).
The efficacy of statins in the primary and secondary preven tion of CVEs has been demonstrated in large scale trials and meta analyses (12). CVE reduction is related to the degree of reduction of low density lipoprotein (LDL) cholesterol levels. Each millimole per liter reduction in LDL cholesterol is associated with a 20-22% reduction in the risk of myocardial infarction (MI), revascularization, and stroke (12). In RA, high grade inflammation is associated with suppression of total cholesterol, LDL cholesterol, and high density lipoprotein (HDL) cholesterol levels, as well as changes in lipid struc ture and function, promoting atherosclerosis (13,14). The potential pleiotropic antiinflammatory/immunomodulatory effects of statins (15) may therefore be more relevant in RA than in the general pop ulation. In the Trial of Atorvastatin in Rheumatoid Arthritis (TARA), atorvastatin 40 mg daily, as an adjunct to disease modifying anti rheumatic drug (DMARD) therapy, provided a modest additional benefit for control of inflammation in RA, at least in a subgroup of patients (16), while the Tayside controlled study of rosuvastatin in RA suggested a potentially beneficial effect on C reactive protein (CRP) levels (17). The extent to which statins affect lipid levels and reduce CVEs in RA remains uncertain, due to the small number of RA patients included in general population trials (18).
The lack of robust primary prevention data, coupled with the multifaceted pharmacologic potential of statins in RA suspected at that time, prompted the Trial of Atorvastatin for the Primary Prevention of Cardiovascular Events in Patients with Rheumatoid Arthritis (TRACE RA), the only statin trial with hard CVE end points in this population.  (19), were >50 years of age or had an RA disease duration of >10 years, and gave informed consent. Patients taking statins and those with known CVD requiring statins, DM, myopathy, or other contrain dications to statins were excluded (see Supplementary Methods 1). Recruiting centers continued their routine practice for screen ing (or not) for cardiovascular risk. There were no restrictions with regard to RA treatment prior to or during the trial period, other than the requirement that patients receive stable doses of anti rheumatic medication for the 3 months prior to inclusion in the study. Potentially eligible patients were identified during routine clinical visits, given the patient information sheet, and invited to contact the local trial team if they were interested in participat ing. A screening visit was then arranged. All patients recruited provided written informed consent.

PATIENTS AND METHODS
Randomization and masking. Trial medication was pro vided by Pfizer UK, bottled by an independent pharmaceutical company (Catalent Pharma Solutions UK) to good manufacturing practice standards, and dispensed by the local study pharmacist. The randomization process was incorporated into the drug label ing. Center was the only stratifying variable. Catalent performed the randomization, labeled each bottle with a unique number, and sup plied the packaged drugs to hospital pharmacies with scratch cards to allow a patient's treatment allocation to be revealed, if necessary. On entering the trial, each patient was given a filled and labeled bottle, coded with a unique study number, which was used for all future supplies for that patient. Study treatment remained double blind for patients, investigators, and study personnel throughout.
Procedures. The trial comprised 3 stages: 1) a screening visit to confirm patient eligibility, secure consent, counsel (ver bally and with a leaflet) patients on modifiable cardiovascular risk factors, and randomize patients (there was no run in period); 2) a 3 month visit to check drug tolerability and safety (by liver function tests and creatine kinase [CK] level); and 3) an intended minimum 5 year treatment period. At the screening visit, base line information on demographic characteristics, medical history, family history of premature CVD, smoking status, and concom itant medication was obtained through interview and case note review. The presence of hypertension was assessed by the case report form question, "Is the patient known to have hypertension?" | 1439 Height, weight, and waist circumference were measured. RA dis ease activity, severity (physical function), and quality of life were assessed using the Disease Activity Score in 28 joints (DAS28) (20), the UK version of the Health Assessment Questionnaire (HAQ) (21) disability index (DI), and the EuroQol 5 domain (EQ 5D) instrument (22), respectively. Blood samples were collected for routine measurements of hematologic features, biochemical features, erythrocyte sedimentation rate, CRP, rheumatoid factor, and/or anti-cyclic citrullinated peptide antibodies.
The protocol did not require measurement of lipid levels at baseline. The results of any lipid measurements that had been requested routinely in primary or secondary care over the previous 12 months were recorded in the clinical trial record. If there was more than 1 lipid measurement, the most recent was used. Gen eral practitioners (GPs) were informed if patients were found to have hypertension, DM, or an existing indication for statins (e.g., known hyperlipidemia, DM, or previously known high CVD risk [according to standard guidelines] requiring a statin for primary or secondary prevention).
Randomized patients were followed up at 3 and 6 months and every 6 months thereafter in person or by telephone. Infor mation on trial efficacy and safety end points, disease activity, severity, and concomitant medication was collected at each visit. Patients were asked if they had taken "most," "some," or "none" of their tablets. Patients were considered compliant if they reported taking "most" of their study tablets since their last visit. Study drug administration could be paused, if necessary, for up to 4 weeks without violating the protocol. Patients who experienced a validated primary end point had no further trial visits but were followed up for mortality via linkage with national death registers. Secondary prevention in these patients was decided by the GP and/or treating physician. Patients who were withdrawn from the study for reasons other than a primary end point continued to attend follow up visits to facilitate adverse event and clinical end point data collection. Patients who developed a clinical need for a statin, other than a primary end point, after randomization could be prescribed up to 40 mg of atorvastatin in addition to the ran domized trial medication and remain in the trial.
Outcome measures. The prespecified primary end point was "major vascular events," defined as nonfatal MI, nonfatal pre sumed ischemic stroke, transient ischemic attack (TIA), any cor onary or non coronary revascularization, or cardiovascular death, excluding both confirmed cerebral hemorrhage (International Sta tistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD 10] codes I64-I99) (23) and non coronary cardiac death (ICD 10 codes I00-I15 and I26-I52), occurring dur ing the scheduled treatment period. Secondary end points were the separate components of the primary end point. Tertiary end points included total and cause specific mortality (coronary, other vascular, and nonvascular death separately); hospitalizations; statin safety related outcomes (persistent elevation of alanine transaminase [ALT] or aspartate transaminase [AST] or myopathy [muscle symptoms plus CK >10× the upper limit of normal (ULN)]; and between group differences at study end in lipid levels and health related outcomes (physical function and quality of life).
Additional information about all potential primary end points was collected from medical records, death certificates, and post mortem examinations (where available). An independent trial end points committee reviewed such information on all potential CVEs and deaths and classified them according to prespecified criteria (see Supplementary Methods 1). Information about hos pital admissions was ascertained via linkage, using each patient's unique National Health Service number, with the Health and Social Care Information Centre (HSCIC) for England and Wales and the Scottish Office's Information and Statistics Division (ISD), and the local hospital medical records departments. Information on mor tality and cause of death was obtained via linkage with the HSCIC and ISD. Patients were asked at each visit about adverse events including muscle pain, and ALT, AST, and CK were measured at 3 months. Liver function tests were also performed regularly (usually every 2-3 months) as part of routine DMARD monitoring. At the final study visit, patients were asked to provide a blood sample for lipid and CRP analysis. These samples were shipped to a single laboratory and measured, blinded with regard to treatment group, on an automated validated platform (c311; Roche Diagnostics) using the manufacturer's calibrators and quality control material. Between run coefficients of variation were all <5.2%. LDL choles terol level was estimated using the Friedewald equation (24). However, the ongoing much lower than anticipated CVE rate led to premature termination of the trial.
All randomized patients were included in the analysis up to December 20, 2012 or the end of trial visit, whichever was ear lier, irrespective of whether the study drug was continued (i.e., intent to treat analyses). Cox regression models were devel oped for time to occurrence of a first CVE using treatment allo cation as the independent variable. The models were stratified by center and adjusted for baseline imbalances, compliance, and nonstudy statin use (the latter two as time dependent vari ables using a previously described method [25]). All adjustments were prespecified in the protocol and the statistical analysis plan (Supplementary Methods 4, available on the Arthritis & Rheumatology web site at http://onlin elibr ary.wiley.com/doi/10.1002/ art.40892/ abstract). Treatment differences were expressed as hazard ratios (HRs) with 95% confidence intervals (95% CIs). P values less than 0.05 (2 sided) were considered significant. Kaplan Meier product limit estimates of the survival curves were calculated. Lipid levels and levels of blood tests monitoring stat in safety were compared between groups using Mann Whitney tests. All analyses were performed with SPSS Statistics for Win dows, version 22.0 (IBM).

RESULTS
Trial progress. Between August 7, 2007 and Novem ber 21, 2011, 3,002 patients with RA from 102 centers were randomized (1,504 to receive atorvastatin and 1,498 to receive placebo). Their mean age was 61 years (228 [7.6%] of 3,002 patients were <50 years of age), and 74% were women. They were followed up for a median of 2.51 years (interquartile range [IQR] 1.90, 3.49 years), providing 7,827 person years of fol low up. At the time of trial closure (December 31, 2011), the observed event rate in the 2 arms combined was 0.70% per annum compared with the expected 1.6% per annum. Trial prog ress is shown in Figure 1 and Supplementary  (2) ACE inhibitors 10/113 (9) 10/127 (8) Other cardiac drugs 10/113 (9) 10/123 (8) * The variable number of patients for each characteristic is due to missing data from incomplete case report forms. The low number of baseline lipid measurements is because, due to budgetary constraints, the protocol did not require measurement of lipid levels at baseline. In the UK, it is the responsibility of primary care physicians to assess their patients for cardiovascular risk and to prescribe statins for primary prevention where indicated according to national guidelines. The  Baseline characteristics of the patients. At base line, the randomized groups were well balanced for all demographic, anthropometric, and RA characteristics, anti rheumatic and other therapies, and CVD risk factors, except for current smoking and nonsteroidal antiinflammatory drug (NSAID) or cyclooxygenase 2 inhibitor (coxib) treatment, which were higher in the atorvastatin group than in the placebo group (18.4% versus 14.5% for current smoking and 42.9% versus 38.0% for NSAID or coxib treatment). A total of 40.3% of the patients had low disease activity according to the DAS28 or were in remission (DAS28 ≤3.2), 86% were receiving stable doses of conventional synthetic DMARDs, 16% were receiv ing biologic DMARDs, and 17% were receiving steroid therapy (Table 1 and Supplementary Table 2 Compliance and nonstudy statin use. In the atorva statin group, reported compliance fell from 89% at the 3 month visit to 39% by 60 months of follow up, while nonstudy statin use increased from 0.5% to 5.6%. In the placebo group, com pliance fell from 89% to 25% and nonstudy statin use increased from 0.7% to 7.8% (Supplementary Table 3 ; P = 0.115). After adjustment for baseline dif ferences, compliance, and nonstudy statin use, the HR was 0.60 (95% CI 0.32, 1.15) (P = 0.127). Based on the number of events, the numbers of patients, and the mean follow up time in each arm, the number needed to treat to prevent 1 CVE during the trial was 121. Kaplan Meier analysis of time to primary end point in the 2 groups is shown in Figure 2, and the cumulative incidence of first CVE in the 2 groups is shown in Figure 3   Life threatening but nonfatal serious adverse events (SAEs) occurred in 22 patients in the atorvastatin group versus 24 patients in the placebo group, while SAEs resulting in death occurred in 19 and 18 patients, respectively. None were considered to be related to trial medication. The randomization code was broken at the site in 3 cases in the atorvastatin arm (due to liver cysts with elevated ALT level, acute hepatitis, and abnormal findings on liver function tests) and 2 cases in the placebo arm (due to high grade lym phoma and chest infection), none of which was attributed to trial medication.
There were 64 reports of "RA flare" (significant worsening of RA symptoms), 29 in the atorvastatin group and 35 in the placebo group. There were 249 reports of "new or significant muscle pain," 132 in the atorvastatin group versus 117 in the placebo group (P = 0.354). Of these, 13 (9 patients in the atorvastatin group versus 4 patients in the placebo group) had concurrent ALT or AST eleva tion of >2× the ULN, which was neither sustained nor considered to be related to the trial medication. Three of these patients (2 in the atorvastatin group and 1 in the placebo group) were withdrawn from the trial by the local principal investigator. None of the patients had a CK elevation of >10× the ULN. Two patients (1 in each group) had unsustained CK elevations of 3-10× the ULN; neither was considered to be due to trial medication. No asymptomatic cases of CK elevation were detected either on monitoring per protocol or during routine DMARD monitoring. There were no cases of ALT or AST elevation of >5× the ULN on per protocol testing, but there were 6 cases outside protocol testing (all unsustained and consid ered unrelated to trial medication). Overall, there were 159 cases of ALT or AST elevation of 2-5× the ULN (90 in the atorvastatin group and 69 in the placebo group; P = 0.103), none of which was sustained or attributed to trial medication.
Biochemical and arthritis outcomes at the end of the trial. At the end of the trial, mean LDL cholesterol levels were 0.77 mmoles/liter lower among those allocated to receive atorva statin compared to those allocated to receive placebo. In the atorvastatin group, 54% of the patients were classified as compliant at the end of trial visit. There were no significant differences between groups in RA disease activity (DAS28), severity (HAQ DI), or quality of life (EQ 5D). However, CRP levels were significantly lower in the ator vastatin group (median 2.59 mg/liter [IQR 0.94, 6.08]) than in the placebo group (median 3.60 mg/liter [IQR 1.47, 7.49]) (P < 0.0001). Although levels of CK and ALT (but not AST) were statistically sig nificantly higher (by ~12-15%) in the atorvastatin group (Table 3), these differences are not considered clinically significant. The num ber of cases of myopathy and elevations of liver enzyme levels above the normal range were similar in the 2 groups. In the end of trial analysis of the atorvastatin group, lipid, ALT, and CRP levels were significantly associated with compliance (data not shown).

DISCUSSION
TRACE RA was designed to assess whether patients with RA who were not already receiving statin therapy would ben efit from atorvastatin 40 mg daily for the primary prevention of CVEs. In this study, the largest ever academically led clinical trial in RA, >3,000 RA patients were recruited and followed up for a median of 2.5 years. The unexpectedly low event rate and resulting limited statistical power to detect an effect during the planned 5 years of follow up led to premature termination of the trial. The best estimate of the "true" reduction in CVEs in the atorvastatin versus placebo arm is 34%. Using a 95% confidence level we cannot rule out any effect size between a 61% reduction and an 11% increase. Thus, our results were not statistically significant. The observed 34% reduction is consistent with the Cholesterol Treatment Trialists' Collabora tion meta analysis of the effect of statins in other populations (12). Furthermore, in this potentially vulnerable population, atorvastatin was safe, with no excess reports of muscle pain or other significant symptoms among those allocated atorva statin compared to those receiving placebo.
There were several reasons such a trial was needed. CVD remains a major cause of death (5) and is significantly increased in people with RA compared to the general population (26), a fact recognized by the addition of RA as an independent risk factor in CVD risk algorithms such as QRISK2 (27) and QRISK3 (28). The relative contribution of classic CVD risk factors and novel mecha nisms related to systemic inflammation to the excess CVD mortal ity of RA is still debated (4,29,30), and there have been no clinical end point trials assessing the effect of statins, or any other primary prevention strategy, in this population. Some small studies have shown that statins reduce surrogate measures of atheroscle rotic events, for example, arterial stiffness (31) or carotid plaque (32), while a few cohort studies have suggested that statin use is associated with survival gains (33) and statin discontinuation with poorer survival (34) in RA. Finally, post hoc analyses of two trials of more intensive versus standard statin doses have suggested that the effect of statins, in terms of both LDL cholesterol reduc tion and CVE prevention, is similar in subjects with "inflammatory joint disease," including RA, and those without joint inflammation; however these findings were based on a very small number of patients and events (18).
Randomization in TRACE RA was stratified only by study site in the expectation that, given the large numbers, baseline variables would distribute evenly between the treatment arms. However, baseline current smoking and NSAID/coxib usage, both well established risk factors for CVEs (35,36), were higher in the atorvastatin group. Although every effort was made to maximize adherence to the trial medication during the trial, adherence rates in TRACE RA were relatively low. Adherence to trial medication in statin trials varies widely (37). This appears to depend on many factors, including the population studied, whether it is for primary or secondary prevention, trial design (e.g. inclusion of a "run in" period), trial duration, and method of assessing adherence, among others. Adherence to statin treatment in real world use is generally accepted to be <50% (37,38). In this context, the adher  (39) and polypharmacy (40), often with potentially hepatotoxic drugs. Virtually all participants in TRACE RA were receiving potentially hepatotoxic therapies such as methotrex ate, but all patients receiving methotrexate were also prescribed folic acid. The 40 mg daily dose of atorvastatin is also of inter est, as there are few randomized data on safety for this dose. Reassuringly, the type and severity of adverse events, the rate of hospitalizations, elevations of liver or muscle enzyme levels, inci dence of myalgia, and worsening of RA were all similar in the 2 arms. These results suggest that atorvastatin 40 mg (and lower doses) is safe to use in patients with RA who are already receiving DMARD therapy.
Clinically assessed RA disease activity, severity, and quality of life were not significantly different between the 2 groups at the end of the trial. However, consistent with data from other studies (16,17), levels of CRP were significantly lower, by ~1 mg/liter, in the atorvastatin group than in the placebo group. This difference is unlikely to be clinically significant in the context of RA disease activity.
Since TRACE RA was terminated early because the CVE rate was much lower than expected, it is not surprising that the HR for the primary end point was not significant. The observed number of primary outcomes provides <20% power to detect the relative risk reduction of 32% specified in the final protocol and provides adequate power (>80%) only to detect a relative risk reduction of >68%. The results for the pri mary outcome are therefore best represented as the estimated HR and its associated confidence interval. When the trial was designed (2002)(2003)(2004), the assumption of a 1.6-1.8% annual event rate seemed, if anything, conservative. A meta analysis of mortality studies in RA published prior to 2005 demon strated a meta-standardized mortality ratio of 1.5 (1). Annual CVE rates ranged from 2.5-5% (26). Possible explanations for the lowerthanexpected observed event rate in TRACE RA include: 1) event rates in randomized trials are always lower than in observational studies and the "healthy volunteer" effect may have been more pronounced than usual; 2) TRACE RA, by design, excluded patients with the highest baseline CVE risk since these patients were already being treated or had a recommendation for a statin; 3) TRACE RA participants were younger than in other statin trials and were predominantly female (as expected from RA disease demographics); and 4) <20% of participants had high disease activity at baseline. There is increasing evidence that good disease control reduces the progression of subclinical atherosclerosis in RA patients (41,42) and is associated with better cardiovascular out comes. Therefore, an additional explanation for the low event rate observed in TRACE RA might have been a significant increase in the use of DMARDs, particularly biologic DMARDs, during the course of the trial. However, this was not the case. The use of prednisolone (in terms of frequency and average daily dosage), conventional synthetic DMARDs, and biologic DMARDs at base line was balanced between the atorvastatin and placebo groups and remained so during the trial. There was not any significant increase in the use of biologic DMARDs during the trial in either group (Supplementary Section 1, available on the Arthritis & Rheumatology web site at http://onlin elibr ary.wiley.com/doi/10.1002/ art.40892/ abstract). Some recent studies suggest a modest decline in CVE rates, mirroring those observed in the general population (43,44), while others demonstrate a very substantial decline in CVE rates in RA (45). Although it is possible that CVEs may have been missed in both the atorvastatin and placebo arms, we believe this is unlikely due to the robust CVE capture system including regular patient review (80% of the patients attended the end of trial visit in both arms) and linkage with several national electronic data sources. Information from data linkage was avail able for all patients.
Overall, the findings of TRACE RA have important implications for clinical practice and research. The large randomized statin trials have shown that statin therapy reduces CVE risk by approximately one third, regardless of the level of background risk. Neverthe less, most guidelines recommend therapy only for those whose estimated individual 10 year or lifetime risk falls above a certain threshold, for reasons of cost and risk/benefit ratio (46). TRACE RA suggests that contemporary RA patients are likely to derive the same level of benefit from statins as other populations. However, the low event rate shows that there is a sizeable population of RA patients (even among those older than 50 years or with >10 years of disease duration) who have a relatively low CVD risk. This finding does not support prescribing statins to all RA patients, one of the main questions addressed by this trial. Instead, the decision to prescribe should be based on assessment of the individual RA patient's risk using, at present, the relevant national or inter national recommendations and risk assessment tools (9), while disease specific algorithms are developed and validated (47). In terms of future research, TRACE RA provides information about effect and sample sizes that may be helpful in the design of future trials investigating CVD prevention strategies in RA, whether these are based on cardiovascular interventions, intensive inflammatory disease control, or both.
In conclusion, TRACE RA suggests that atorvastatin 40 mg daily is safe for the primary prevention of CVEs in patients with RA and appears to confer a similar degree of risk reduction in these patients as in other populations. CVE rates are decreasing in this population. This finding requires further investigation and does not support a pri mary prevention strategy involving statin use in all RA patients.