Author + information
- Received September 27, 2012
- Revision received February 12, 2013
- Accepted February 19, 2013
- Published online August 20, 2013.
- Peter S. Sever, MB BChir, PhD∗∗ (, )
- Neil R. Poulter, MB BS∗,
- Choon L. Chang, PhD∗,
- Simon A.M. Thom, MD∗,
- Alun D. Hughes, MB BS∗,
- Paul Welsh, PhD†,
- Naveed Sattar, MB BS, PhD†,
- ASCOT Investigators
- ∗International Centre for Circulatory Health, National Heart & Lung Institute, Imperial College London, London, United Kingdom
- †Institute of Cardiovascular and Medical Sciences, University of Glasgow, Glasgow, United Kingdom
Reprint requests and correspondence:
Professor Peter S. Sever, International Centre for Circulatory Health, Imperial College London, 59 North Wharf Road, London W2 1LA, United Kingdom.
Objectives The aim of this study was to determine whether baseline and on-statin C-reactive protein (CRP) are independent predictors of cardiovascular (CV) outcome beyond low-density lipoprotein cholesterol (LDL-C).
Background Use of CRP as a predictor of statin treatment remains controversial.
Methods We investigated the relationship of baseline and on-treatment CRP with subsequent CV events in Cox models using a subset of white subjects with no history of CV disease from the UK ASCOT (Anglo-Scandinavian Cardiac Outcomes Trial).
Results During 5.5 years of follow-up, a total of 488 subjects experienced a CV event. CV risk increased with baseline CRP (hazard ratio [HR] per 1 SD: 1.21; 95% confidence interval [CI]: 1.09 to 1.33) in an adjusted model. In ASCOT Lipid-Lowering Arm, the relative statin effect in preventing CV events did not differ according to tertiles of baseline CRP (p = 0.69). After 6 months of atorvastatin therapy, the median LDL-C and CRP were reduced by 38.7% and 25.8%, respectively. Those who achieved LDL-C below the median had a reduced CV risk (HR: 0.58; 95% CI: 0.34 to 0.97) compared with those who did not. In contrast, those who achieved a CRP level below the median did not have a reduced risk of CV events (HR: 0.95; 95% CI: 0.59 to 1.55). Among those who achieved LDL-C below the median, there was no difference in CV risk whether they also achieved a CRP level below (HR: 0.55; 95% CI: 0.30 to 1.02) or above the median (HR: 0.56; 95% CI: 0.30 to 1.03).
Conclusions In these primary prevention patients, although baseline CRP independently predicted CV event risk, the achieved CRP level on while statin therapy did not predict CV events, either alone or in combination with LDL-C.
That inflammation plays an important role in the pathophysiology of atherosclerosis is undisputed (1–3). Moreover, various molecular and cellular components of the inflammatory response are activated and may contribute to plaque rupture and the presentation of acute coronary syndromes (4). Therefore, the study of biomarkers of inflammation, notably C-reactive protein (CRP) and serum amyloid A, has been a focus of interest for many authors (5,6). Indeed CRP has been proposed, and in some places used, as a clinical tool for cardiovascular disease (CVD) risk prediction scores (7–9), yet there remains disagreement regarding its clinical utility, beyond that achieved with conventional biomarkers (10). There has also been notable debate over whether statin-associated CRP reduction is an independent predictor of CVD after consideration of concomitant LDL-cholesterol (LDL-C) reduction (11,12).
We previously reported, using a nested case-control design, from the ASCOT (Anglo-Scandinavian Cardiac Outcomes Trial) in hypertensive patients selected on the basis of traditional risk factors, that CRP did not usefully improve the prediction of CV events and, critically, that reduction in CRP associated with statin therapy was not a predictor of CV outcome alone or in combination with LDL-C (12). One criticism of our findings was based on the relatively small number of patients in whom on-treatment levels of CRP were available (11), partly a consequence of the matched case-control methodology. Therefore, to address these concerns by increasing statistical power and to limit potential bias, we have now extended our observations to include a full cohort analysis of all eligible white patients in the United Kingdom (UK) and Ireland recruited into the ASCOT-LLA (Lipid-Lowering Arm of the Anglo-Scandinavian Cardiac Outcomes Trial) from whom baseline and on-treatment values of CRP were obtained. Our hypothesis was that on-treatment CRP would not meaningfully predict subsequent CVD events, whereas LDL-C was strongly and significantly predictive.
A cohort study based on subjects recruited into the ASCOT-LLA cohort was used to determine the association between baseline CRP and subsequent CV outcomes. In separate analyses, we assessed whether CRP levels after 6 months of treatment with atorvastatin 10 mg were independent predictors of CV outcomes. For the purpose of the present study, those with any history of CVD were excluded.
Patients and recruitment
The detailed ASCOT protocol was published previously (13), and further information is available at http://www.ascotstudy.org. Hypertensive patients with ≥3 other risk factors for CVD but no history of myocardial infarction (MI) or currently treated angina were eligible.
In the ASCOT-BPLA (Blood Pressure-Lowering Arm of the ASCOT), 9,098 patients were randomized in the UK and Ireland to either amlodipine, adding perindopril as required (amlodipine-based), or atenolol, adding bendroflumethiazide as required (atenolol-based).
In addition to randomization into the ASCOT-BPLA, those with a fasting total cholesterol of ≤6.5 mmol/l (250 mg/dl) were further randomized using a factorial design to either 10 mg/day atorvastatin or matching placebo (ASCOT-LLA).
The ASCOT-LLA was terminated prematurely after a median follow-up of 3.3 years owing to highly significant benefits in favor of atorvastatin over placebo on the primary coronary endpoint. All patients in the ASCOT-LLA were offered open-label atorvastatin and continued in the ASCOT-BPLA until its termination after a median 5.5 years of follow-up.
Baseline characteristics of participants and primary outcomes of each arm of the trial were previously reported (13–15). In the current analyses, only white patients with no history of CVD were included from both the ASCOT-LLA and BPLA combined.
The major CV outcome evaluated in analyses was a composite of fatal coronary heart disease (CHD), symptomatic nonfatal MI, coronary revascularization, fatal and nonfatal stroke occurring in the UK and Ireland among participants in the ASCOT-LLA study between February 1998 and October 2005. During the median follow-up period of 5.5 years, 488 CV outcomes were reported. In addition, associations with CHD events (fatal CHD, symptomatic nonfatal MI, coronary revascularization), and stroke events were also investigated where numbers permitted.
Fasting lipids were routinely measured annually during the ASCOT-LLA trial. CRP samples were collected at baseline and after 6 months, and subsequently all stored serum samples were measured by a high-sensitivity method, on an Abbott Architect (Abbott Diagnostics, Abbott Park, Illinois), by technicians blinded to the CV outcome status of the participants’ samples. The lower limit of sensitivity was 0.1 mg/l and the coefficient of variation was <4%.
Continuous data are presented as mean ± SD or median and interquartile range (IQR) and categorical variables as proportions. Baseline characteristics were compared by event or no-event groups using t tests, chi-square tests for proportions, and nonparametric Mann-Whitney tests for skewed data. Age- and sex-adjusted Spearman correlation was used to assess the correlation between baseline CRP and baseline clinical characteristics. For a given sample size of 3,987 subjects with 456 CV events, the study had at least 90% power to detect a hazard ratio (HR) of 1.20 per SD increased in loge CRP. Analyses were performed on an intention-to-treat basis, and person-time was calculated until the first confirmed CV endpoint or to the end of the trial if no endpoint occurred. Kaplan-Meier curves were used to estimate cumulative incidence over time by baseline CRP tertile groups, and the log-rank test was used to compare survival curves. The association between CRP and the risk of each of CHD or stroke event was reported as an HR obtained from a Cox proportional hazard regression model, first by treating loge-transformed baseline CRP as a continuous variable giving the risk of having an event per 1-SD change with 95% confidence intervals (CIs) and, second, by categorizing CRP into tertiles with the lowest as a reference. Two models were used to examine the association between baseline CRP and the risk of CVD: model 1, adjusted for age and sex; model 2, adjusted for age, sex, current smoking status, diabetes mellitus, left ventricular hypertrophy, baseline systolic blood pressure, total and high-density lipoprotein cholesterol (HDL-C), randomized statin drug assignment, randomized blood pressure drug assignment, body mass index (BMI), fasting glucose, family history of CHD, creatinine, and educational attainment. However, 2 different models were used in the on-treatment CRP analyses: model 1, adjusted for age, sex, and loge baseline CRP, and model 2, adjusted for age, sex, current smoking status, diabetes mellitus, left ventricular hypertrophy, baseline systolic blood pressure, total and HDL-C, randomized atorvastatin/placebo, randomized blood pressure drug assignment, BMI, fasting glucose, family history of CHD, creatinine, educational attainment, loge baseline CRP, and LDL-C. The assumption of proportionality was tested with Schoenfield’s residuals. To test for effect modification by CRP tertiles, we included interaction terms between CRP or LDL-C tertile indicators and randomized assignment in the models. Because of the small number of subjects (n = 47; 26 receiving placebo and 21 receiving atorvastatin) with stroke outcomes in the ASCOT-LLA, with complete data on CRP, LDL-C, and covariates, we only reported the on-treatment effect on CVD and CHD.
To investigate the predictive effect of baseline CRP and LDL-C on CVD and CHD, analyses used UK ASCOT-LLA and BPLA combined datasets, whereas for the on-treatment analyses, we restricted analyses to those who participated in the ASCOT-LLA. On treatment CRP/LDL-C levels were treated as time-dependent covariates in these analyses.
Two sensitivity analyses were performed to assess the consistency of our results. Analysis of the association between baseline CRP and each endpoint was repeated on those who were not assigned to receive atorvastatin. The second used imputed data for missing data values to repeat on-treatment CRP analyses.
Ten imputed datasets were created using Rubin’s rules (16) to combine effect estimates and estimate SEs to allow for the uncertainty caused by missing data. Multiple imputation allows inclusion of subjects with incomplete data in analyses and makes full use of all the available data, increasing power and precision. Multiple imputation was undertaken using the imputation by chained equation in Stata version 11.0 (StataCorp, College Station, Texas).
We further performed a post hoc analysis to investigate the association between on-treatment non-HDL-C and CV and CHD events.
Analyses were conducted using SAS version 9.1 (SAS Institute, Cary, North Carolina) and Stata version 11.0 (StataCorp), using 2-sided tests with a significance level of <0.05.
In total, 488 subjects with an event and 3,797 with no event during follow-up in the ASCOT-LLA and ASCOT-BPLA were included in the analyses of predictive values of baseline CRP (Fig. 1). The mean age of the subjects was 64.3 ± 8.1 years, and 85.6% were male. A comparison of the baseline characteristics between those in whom an event developed and those in whom one did not demonstrated that those in whom an event developed had a generally worse clinical profile (Table 1).
Baseline CRP was positively correlated with total cholesterol, LDL-C, triglyceride level, and BMI but negatively with HDL-C in both groups (Online Table 1).
Baseline CRP and risk of subsequent CV events
The risk of the development of a CV event and CHD alone increased with baseline CRP (Table 2). Baseline CRP was not, however, significantly associated with the risk of stroke alone, although there were trends toward a positive association. The HR for CV events was 1.21 (95% CI: 1.09 to 1.33; p = 0.0003) per 1 SD increased in loge-transformed CRP, after adjusting for classic risk factors and randomization. Similar results were noted in the analyses by tertiles of CRP (Table 2, Fig. 2). Similar results were observed in a sensitivity analysis of subjects who were not assigned to receive atorvastatin (Online Table 2).
Statin efficacy effect by baseline CRP
There was no evidence of an interaction between baseline LDL-C or CRP and treatment effect (statin-/placebo- or atenolol-/amlodipine-based treatment) on CV events or CHD or stroke; specifically, the statin effect in preventing CVD did not differ significantly according to the tertiles of baseline CRP (p > 0.60) (Online Table 3).
Achieved CRP and LDL-C at 6 months and risk of CV event
After 6 months of atorvastatin treatment, the median LDL-C was reduced by 38.7% (3.46 mmol/l [IQR: 2.96 to 3.95 mmol/l] to 2.12 mmol/l [IQR: 1.74 to 2.56 mmol/l]), whereas in the placebo group, the median decreased by 1.7% (from 3.46 mmol/l [IQR: 2.99 to 3.96 mmol/l] to 3.40 mmol/l [IQR: 2.86 to 3.92 mmol/l]; comparing change, p < 0.0001). The concomitant changes for CRP were a 25.8% reduction with atorvastatin (from 2.21 mg/l [IQR: 1.12 to 4.63 mg/l) to 1.64 mg/l (IQR: 0.82 to 3.43 mg/l) compared with 0.4% in the placebo group (from 2.25 mg/l [IQR: 1.09 to 4.4 mg/l] to 2.24 mg/l [IQR: 1.13 to 4.48 mg/l]; p = 0.02). The Spearman correlation between the percentage of change in CRP and the percentage of change in LDL-C was modest (r = 0.12, p < 0.0001).
Changes in LDL-C or CRP from baseline to 6 months were normally distributed. Irrespective of treatment group assignment in the ASCOT-LLA cohort, a 1-mmol/l decrease in LDL-C between baseline and 6 months of in-trial treatment was associated with a 17% (HR: 0.83; 95% CI: 0.69 to 0.99; p = 0.04) and a 17% (HR: 0.83; 95% CI: 0.67 to 1.02; p = 0.07) risk reduction in CV and CHD events, respectively, after adjustment for baseline levels, age, and sex. These estimated effects were essentially unchanged after multiple adjustments for other risk factors, but were no longer significant (HR: 0.86; 95% CI: 0.66 to 1.12 and HR: 0.87; 95% CI: 0.63 to 1.19, respectively). In contrast, the effect of a 10-mg/l decrease in CRP showed no evidence of an association with CV or CHD events (HR: 1.03; 95% CI: 0.90 to 1.18; p = 0.69 for CV and HR: 1.05; 95% CI: 0.92 to 1.20; p = 0.48 for CHD). Those who did not achieve CRP below the median value at 6 months in either the placebo (median, 2.24 mg/l) or atorvastatin (median, 1.64 mg/l) group did not have a significantly altered risk of CV or CHD events compared with those who did (Table 3).
Similar analyses were repeated using a separate imputed dataset. In this dataset, the multivariable analyses (model 4) showed that the HRs for CV were similar to those reported in the complete case analysis of Table 3 (Online Table 4).
After adjusting for risk factors (including baseline CRP and LDL-C), subjects allocated to atorvastatin had a nonsignificant 19% to 22% reduced risk of CV events regardless of whether they achieved a CRP less than the median of 1.64 mg/l (Fig. 3, Table 4). In contrast, those who achieved LDL-C below the median had a significant 46% reduction in the risk of CV events (HR: 0.54; 95% CI: 0.34 to 0.85; p = 0.008) in the multivariable-adjusted analysis. In the atorvastatin group, those who achieved LDL-C below the median had a 42% reduced risk compared with those who did not (HR 0.58; 95% CI: 0.34 to 0.97; p = 0.04) (model 2 in Table 4). Similar results were also noted in the CHD analyses (Fig. 3, Table 4). Post hoc analyses of on-treatment non-HDL showed similar significant results (data not shown).
On-treatment CRP and LDL-C analyses were repeated using a dataset including imputed data for missing data. A significant treatment effect for CHD (39% risk reduction) was observed in those who achieved CRP below the median. However, in contrast to findings for achieved LDL-C, there was, once again, no significant atorvastatin group difference for achieved CRP below, versus above, the median (HR: 0.75; 95% CI: 0.43 to 1.32; p = 0.32). Overall, the effect estimates of on-treatment CRP and LDL-C on CV and CHD events were similar to those reported in the complete case analyses (Table 5).
Compared with placebo, the lowest risk of CV events was noted in subjects allocated to atorvastatin who achieved LDL-C below the median level (<2.1 mmol/l) irrespective of on-treatment CRP levels (Fig. 4, Table 6). A significantly lower risk of CHD was noted in subjects taking atorvastatin who achieved low LDL-C but had a high level of on-treatment CRP (Fig. 5, Table 6). However, this result was based on 6 subjects with an event. Sensitivity analyses using imputed data showed similar results (Fig. 6).
We conducted additional analyses using alternative cutoffs such as CRP more than and less than 2 mg/l and either LDL-C more than and less than 2.59 mmol/l (100 mg/dl) or non-HDL-C more than and less than 3.37 mmol/l (130 mg/dl). In neither of these analyses was there an additional benefit to be gained by lowering CRP to <2 mg/l (data not shown).
We report in this trial of hypertensive patients at modest risk of future CV events that baseline CRP independently predicted subsequent CV events but neither baseline levels of CRP nor the achieved levels of CRP on-treatment with atorvastatin 10 mg/day predicted the efficacy of the statin in preventing future CV events. The results of this cohort analysis are similar but more robust than those that we reported from an earlier nested case-control study derived from the same population (12).
Several studies have shown that baseline CRP is an independent risk predictor of CV events, and this has been confirmed in a recent meta-analysis (6) such that predictive benefits are, at best, modest (9). Although there remains uncertainty about the relative benefit of CRP for risk prediction, the strength of the association, particularly when potential confounders are incorporated into the models, is weak (6). We also confirm in this study that there was no interaction between baseline values of CRP and the relative effect of the statin on CV events. These findings are consistent with those reported from the Heart Protection Study (17), the PROSPER (Prospective Study of Pravastatin in the Elderly at Risk) trial (18), and the JUPITER (Justification for the Use of statins in Prevention: Intervention Trial Evaluating Rosuvastatin) trial (19).
Arguably the most controversial and debatable issue concerns the extent to which on-treatment levels of CRP predict the benefits of CV outcome associated with statin therapy. This has been proposed based on analyses of observations of on-treatment levels of CRP from a number of trials, many of which recruited high-risk individuals (20–22) and, particularly, JUPITER, which recruited low-risk individuals with high CRP (19).
It has been suggested that data from our earlier nested case-control study in the ASCOT-LLA showed that those who achieved lower levels of on-treatment CRP had a 25% greater relative risk reduction in CV events and that our results were actually compatible with those of other studies (11). There are flaws in this argument. First, when controlling for confounding variables including CRP and LDL-C at baseline, the absolute difference in risk was 12% (not 25%) (12). Further, those who did not achieve median CRP reduction had an increased risk of 14% in the ASCOT case-control study, with figures of 5% and 12% increased risk in the analyses using complete case data and imputed data, respectively, in the present study. These estimates appear appreciably different from the 47%, 33%, and 26% increased risk reported in the JUPITER trial (19), A to Z (Aggrastat to Zocor Trial) (21), and PROVE IT–TIMI 22 (Pravastatin or Atorvastatin Evaluation and Infection Therapy–Thrombolysis in Myocardial Infarction 22) trial (20) for similar comparisons. Regardless of treatment group assignment in the ASCOT, a 1-mmol/l decrease in LDL-C was shown to be associated with a significant 17% reduced risk of CV and CHD events. A reduction in CRP, however, did not have any association with CV outcome in the ASCOT-LLA population, once reduction of LDL-C was taken into account. Unless the argument is being made that statins are in fact primarily anti-inflammatory drugs (rather than cholesterol-reducing drugs) selective or even additional reporting of the risk reduction association with CRP lowering appears of minimal benefit in delineating the impact of statins on CV risk. Figure 4 clearly shows that once achieved LDL-C is considered, further stratification by achieved CRP has little bearing on the relative risk of CV events. Because about half the patients recruited into the ASCOT-LLA could be categorized as having the metabolic syndrome, thus sharing some characteristics with the JUPITER trial population, we conducted a post hoc subgroup analysis of these patients. Compared with those in the placebo group, among those with the metabolic syndrome, those who were assigned to receive atorvastatin had a higher on-treatment LDL-C and tended to have higher risk of CVD regardless of their on-treatment CRP levels. However, those with lower on-treatment LDL-C and lower on-treatment CRP had a nonsignificant 35% reduced risk of CVD, but there was no evidence of interaction between the metabolic syndrome and on-treatment CRP and LDL-C levels on CVD. Such analyses should be interpreted with caution, however, owing to the limited number of events in individual subgroups.
A further commentary on our earlier findings from the case-control study was made on the grounds of inadequate power (11). The current study reports on 456 CV events in total, and for the on-treatment CRP analysis in 97 events (with imputed data). This compares with 103 CV events in the JUPITER trial for the on-treatment analysis (19). Moreover, the JUPITER primary CV endpoint definition included hospitalization for angina, which in the ASCOT CV definition was not considered a hard endpoint and therefore not included. We therefore believe that the power of both studies is comparable. We, however, subsequently conducted a further analysis incorporating 11 cases of the development of unstable angina as a CV endpoint. This made no difference to our original conclusion that those achieving lower levels of LDL-C had a lower risk of CV events regardless of the achieved level of CRP.
Our results showing that on-treatment CRP has virtually no predictive value are in accord with the TNT (Treating New Targets) study (23), and CARDS (Collaborative Atorvastatin Diabetes Study) (unpublished data) but contrast with other trials performed on patients with acute coronary syndrome (20,21) and patients with angiographically documented coronary disease (20). Investigators have previously claimed that lower levels of CRP after statin therapy predict greater subsequent relative risk reductions in CV events (19–22). Unfortunately, none of these studies except REVERSAL (Reversal of Atherosclerosis with Aggressive Lipid Lowering) (22) reported the relationship of change in CRP (from baseline to follow-up) with events.
Study limitations and strengths
Strengths and limitations of the present study require consideration. The relative merits of primary endpoints and power of studies reporting on treatment CRP and CV events were discussed earlier. The present study reports objectively imputed data, although these results are entirely consistent with results from nonimputed data. In terms of relative generalizability, the PROVE-IT (20) and A to Z (21) trials recruited high-risk patients with previous acute coronary events. REVERSAL trials recruited patients with at least 20% stenosis on coronary angiography and with a mean baseline LDL of 3.9 mmol/l and geometric mean CRP of 2.9 mg/l (22). The JUPITER trial recruited relatively healthy patients with low LDL-C (<3.4 mmol/l) but high CRP (≥2 mg/l) (19), whereas the ASCOT cohort in these analyses included patients without any history of CV and with a median LDL-C of 3.6 mmol/l and a median CRP of 2.4 mg/l at baseline. It is possible that CRP may have a different predictive ability in different patient types. However, we believe that, to date, ASCOT data most usefully reflects patients commonly seen in primary care.
Our results challenge the need to measure CRP to guide statin dose changes. These clinical decisions appear usefully informed by the change in LDL-C while on statin treatment, in line with the pharmacological role of statins in cholesterol reduction.
For supplemental tables, please see the online version of this article.
The sponsors of the study (Pfizer) had no role in the study design, data collection, data analyses, data interpretation, or writing of the report. The database was held by the ASCOT Executive Committee who had final responsibility for the decision to submit for publication. Drs. Sever, Poulter, Hughes, and Thom are supported by the Biomedical Research Centre Award to Imperial College Healthcare NHS Trust and the BHF Research Centre Excellence Award to Imperial College. Dr. Welsh is supported by BHF fellowship grant FS/10/005/28147. Dr. Sever and Dr. Poulter have served as consultants to, received travel expenses from, and payment for speaking at meetings for and received research funding from Pfizer to cover administrative staffing and analytical costs of the biomarker analyses. Dr. Sattar has received consulting support from MSD, Bristol-Myers Squibb, and AstraZeneca; and research grant support from Pfizer. Dr. Hughes and Dr. Thom have received research grant support from Pfizer. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- body mass index
- confidence interval
- coronary heart disease
- C-reactive protein
- cardiovascular disease
- high-density lipoprotein cholesterol
- hazard ratio
- interquartile range
- low-density lipoprotein cholesterol
- myocardial infarction
- Received September 27, 2012.
- Revision received February 12, 2013.
- Accepted February 19, 2013.
- American College of Cardiology Foundation
- Entman M.L.,
- Michael L.,
- Rossen R.D.,
- et al.
- Libby P.,
- Ridker P.M.,
- Maseri A.
- Ridker P.M.
- Deveraj S.,
- Singh U.,
- Jialal I.
- Braunwald E.
- Sever P.S.,
- Poulter N.R.,
- Chang C.L.,
- et al.
- Sever P.S.,
- Dahlöf B.,
- Poulter N.R.,
- et al.,
- for the ASCOT Investigators
- Sever P.S.,
- Dahlof B.,
- Poulter N.R.,
- et al.,
- for the ASCOT Investigators
- Dahlof B.,
- Sever P.S.,
- Poulter N.R.,
- et al.,
- for the ASCOT Investigators
- Rubin D.B.
- Sattar N.,
- Murray H.M.,
- McConnachie A.,
- et al.,
- for the PROSPER Study Group
- Ridker P.M.,
- Cannon C.P.,
- Morrow D.,
- et al.,
- for the Pravastatin or Atorvastatin Evaluation and Infection Therapy–Thrombolysis in Myocardial Infarction 22 (PROVE IT-TIMI 22) Investigators
- Morrow D.A.,
- de Lemos J.A.,
- Sabatine M.S.,
- et al.
- Arsenault B.J.,
- Barter P.,
- DeMicco D.A.,
- et al.