high-dose statin therapy in patients with stable coronary...

2485

Lipid-lowering is the cornerstone of secondary prevention for vascular events in patients with coronary artery disease

(CAD). Randomized clinical trials evaluating high-dose versus usual-dose statin therapy have shown that intensive lipid lowering results in additional reduction of vascular events.1–6 Despite this high-quality evidence of superior effectiveness, high-dose statin therapy is not unequivocally recommended for all patients with stable CAD.7,8 This reluctance may be explained by concern about high-dose statins causing higher

rates of side-effects, such as myalgia and transaminits.1–3 Moreover, intensive treatment is more costly and requires closer follow-up of patients, because high-dose versus usual-dose statin therapy is associated with a somewhat higher risk of developing diabetes mellitus.9 Therefore, clinicians need to identify patients who will benefit most from high-dose versus usual-dose statin therapy. Guidelines recommend titration of statin-dose, but only in patients whose low-density lipoprotein (LDL) cholesterol level is not at target.7,8 Although

Background—Clinicians need to identify coronary artery disease patients for whom the benefits of high-dose versus usual-dose statin therapy outweigh potential harm. We therefore aimed to develop and validate a model for prediction of the incremental treatment effect of high-dose statins for individual patients in terms of reduction of 5-year absolute risk for myocardial infarction, stroke, coronary death, or cardiac resuscitation.

Methods and Results—Based on data from the Treating to New Targets trial (TNT; n=10 001), a Cox proportional hazards model was developed comprising 13 easy-to-measure clinical predictors: age, sex, smoking, diabetes mellitus, total cholesterol, high-density lipoprotein cholesterol, systolic blood pressure, history of myocardial infarction, coronary artery bypass grafting, congestive heart failure or abdominal aortic aneurysm, glomerular filtration rate, and treatment status (ie, atorvastatin 80 mg or 10 mg). External validation in the Incremental Decrease in End Points Through Aggressive Lipid Lowering trial (IDEAL; n=8888) confirmed adequate goodness-of-fit and calibration, but moderate discrimination (C-statistic, 0.63; 95% confidence interval, 0.62–0.65). Still, among participants of both trials combined, the model identified a group of 11.7% whose predicted 5-year number needed to treat was ≤25 and a group of 41.9% whose predicted needed to treat was ≥50. A decision curve shows that making treatment decisions on the basis of predictions using our model may improve net benefit.

Conclusions—Estimation of the incremental treatment effect of high-dose versus usual-dose statin therapy in individual coronary artery disease patients enables selection of high-risk patients that benefit most from more aggressive therapy.

Clinical Trial Registration—URL: http://www.clinicaltrials.gov. Unique identifiers: NCT00327691 and NCT00159835. (Circulation. 2013;127:2485-2493.)

Key words: coronary disease ◼ forecasting ◼ prevention & control ◼ statins, HMG-CoA ◼ treatment outcome

© 2013 American Heart Association, Inc.

Circulation is available at http://circ.ahajournals.org DOI: 10.1161/CIRCULATIONAHA.112.000712

Received December 24, 2012; accepted May 2, 2013.From the Department of Vascular Medicine, University Medical Center Utrecht, Utrecht, The Netherlands (J.A.N.D., F.L.J.V.); the Department of

Cardiology, Academic Medical Center, Amsterdam, The Netherlands (S.M.B.); the Julius Center for Health Sciences and Primary Care, Utrecht, The Netherlands (Y.v.d.G.); the Department of Vascular Medicine, Academic Medical Center, Amsterdam, The Netherlands (J.J.P.K.); the Health Science Center, State University of New York, Brooklyn, NY (J.C.L.); the Center for Preventive Medicine, Oslo University Hospital, and University of Oslo, Oslo, Norway (T.R.P.); Pfizer Inc, New York, NY (D.A.D.); and the Division of Preventive Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA (P.M.R., N.R.C.).

The online-only Data Supplement is available with this article at http://circ.ahajournals.org/lookup/suppl/doi:10.1161/CIRCULATIONAHA. 112.000712/-/DC1.

Guest editor for this article was Wendy Post, MD, MS.Correspondence to Frank L.J. Visseren, MD, MSc, PhD, Department of Vascular Medicine, University Medical Center Utrecht, P.O. Box 85500, 3508

GA Utrecht, The Netherlands. E-mail [email protected]

High-Dose Statin Therapy in Patients With Stable Coronary Artery Disease

Treating the Right Patients Based on Individualized Prediction of Treatment Effect

Johannes A.N. Dorresteijn, MD, MSc, PhD; S. Matthijs Boekholdt, MD, PhD; Yolanda van der Graaf, MD, PhD; John J.P. Kastelein, MD, PhD; John C. LaRosa, MD;

Terje R. Pedersen, MD, PhD; David A. DeMicco, DPharm; Paul M Ridker, MD, MPH, FACC; Nancy R. Cook, ScD; Frank L.J. Visseren, MD, MSc, PhD

by guest on July 8, 2018http://circ.ahajournals.org/

Dow

nloaded from


Dow

nloaded from


Dow

nloaded from


Dow

nloaded from

http://www.clinicaltrials.gov

http://circ.ahajournals.org/lookup/suppl/doi:10.1161/CIRCULATIONAHA.112.000712/-/DC1

http://circ.ahajournals.org/lookup/suppl/doi:10.1161/CIRCULATIONAHA.112.000712/-/DC1

mailto:[email protected]

http://circ.ahajournals.org/




2486 Circulation June 25, 2013

this approach seems sensible from a pharmacodynamic perspective, meta-analyses show that the effectiveness of statin therapy in terms of reduction of vascular events does not depend on pretreatment LDL-cholesterol level.6 Instead, the absolute cardiovascular risk reduction achieved by high-dose statin therapy is only proportional to the absolute baseline risk. Previously, Deedwania et al10 have indicated that selecting patients with the metabolic syndrome or diabetes mellitus is a simple and effective method for identifying high-risk patients who significantly benefit from high-dose statin therapy. Such risk stratification, however, may be further improved by taking into account additional patient characteristics.

Clinical Perspective on p 2493In the present study we aimed to develop and validate a model,

based on multiple clinical patient characteristics, for prediction of treatment effect (ie, 5-year absolute risk reduction of major cardiovascular events [MCVEs]) of high-dose statin therapy versus usual-dose statin therapy in individual patients with sta-ble CAD. For this purpose we use data from the Treating to New Targets (TNT) trial and the Incremental Decrease in End Points Through Aggressive Lipid Lowering (IDEAL) trial.1,2 Because such a predicted treatment effect may be used for guiding indi-vidual treatment decisions, we also compare the net benefit of a number of strategies to select high-risk patients with stable CAD to qualify for high-dose statin therapy (ie, [1] all patients, [2] no patients, [3] patients with the metabolic syndrome or dia-betes mellitus, and [4] patients whose predicted treatment effect exceeds a certain threshold). Here we present our results.

MethodsThe design, rationale, and outcomes of the TNT and IDEAL trials have been described in detail elsewhere.1,2,11,12 In brief, both trials evaluated the efficacy and safety of high-dose statin therapy (ie, ator-vastatin 80 mg) versus usual-dose statin therapy (ie, atorvastatin 10 mg in TNT and simvastatin 20 or 40 mg in IDEAL) for the prevention of new vascular events in patients with stable CAD. The TNT trial enrolled 10 001 patients in the United States, Europe, and Australia with a history of myocardial infarction, angina with objective evidence of atherosclerotic CAD, or a history of coronary revascularization. After a median follow-up period of 4.9 years, the hazard ratio for the occurrence of MCVE (ie, fatal or nonfatal myocardial infarction, fatal or nonfatal stroke, fatal coronary heart disease, or resuscitation after cardiac arrest) was 0.78 (95% confidence interval [CI], 0.69–0.89) favoring atorvastatin 80 mg. The IDEAL trial enrolled 8888 patients in Northern Europe with a history of myocardial infarction. After a median follow-up of 4.8 years, the hazard ratio for the occurrence of MCVE was 0.87 (95% CI, 0.78–0.98), favoring atorvastatin 80 mg.

Model DerivationBased on data from the TNT trial a Cox proportional hazards model was developed for prediction of 5-year treatment effect (ie, absolute reduction of MCVE risk) of high-dose statin therapy versus usual-dose statin therapy in individual patients with stable CAD using previously described methods.13,14 Data from the IDEAL trial were reserved as a validation data set. MCVE was defined by the above described primary end point definition of the TNT trial. Candidate predictors were based on the Framingham risk score for general car-diovascular disease,15 with the addition of a few readily available clinical characteristics of particular importance in secondary preven-tion. As a result, the following candidate predictors were considered: age, sex, current smoking, diabetes mellitus, total cholesterol, high-density lipoprotein cholesterol, systolic blood pressure, a history of myocardial infarction, percutaneous coronary intervention, coronary

artery bypass grafting, congestive heart failure (New York Heart Association [NYHA] class I–IIIa) or abdominal aortic aneurysm, glomerular filtration rate (estimated using the Modification of Diet in Renal Disease [MDRD] formula),16 and treatment status (ie, atorvas-tatin 80 mg or usual-dose statin). In accordance with the Framingham risk score, an interaction variable between systolic blood pressure and blood pressure treatment (yes/no) was considered. Finally, because separate Framingham risk scores exist for men and women, any sex-based interactions were evaluated to allow the effect of risk factors to differ between men and women. Interaction terms with treatment were not evaluated because modification of relative treatment effect by clinical patient characteristics was not anticipated based on previ-ous studies.6

One or more covariate data were missing in 43 (0.4%) TNT trial participants, and these were single imputed by weighted probability matching on the basis of multivariable regression using covariate and outcome data, because complete case analysis leads to loss of statisti-cal power and possibly to bias.17,18 Continuous predictors were trun-cated at the 1st and 99th percentile to limit the effect of outliers.18,19 If the association between a continuous predictor and the outcome was not linear, the predictor was transformed to improve model fit.18,19 As a result, a quadratic term was added for age and glomerular filtration rate to obtain optimal model fit.

First, sex-based interactions with probability values ≥0.05 were removed from the starting model. Next, the model was further simplified by stepwise backward selection on the basis of the Akaike’s Information Criterion (AIC). Although all candidate predictors are well-known risk factors for vascular events and, thus, little optimism was expected, the coefficients of the remaining predictors were penalized by uniform shrinkage. The percentage optimism and appropriate shrinkage factor were determined by bootstrap resampling.18,19 Models were fitted for prediction of 5-year MCVE risk. This duration of follow-up was achieved by 4755 TNT participants (47.5%). The proportional hazard assumption was assessed by testing the correlations between scaled Schoenfeld residuals for the various predictors and time.

Model ValidationOne or more covariate data were missing in 248 (2.8%) IDEAL participants, and these were single imputed using the same methods described above. Likewise, continuous variables were truncated at the 1st and 99th percentile. The discriminatory ability of the model (ie, the extent to which the model can separate those with and without a MCVE) was expressed by the C statistic.18 Model calibration, reflecting the precision of how close the predicted survival probabilities are to the actual (observed) survival, was demonstrated by calibration plots.18,19 In addition, overall goodness-of-fit of both models was assessed in the validation cohort with the Gronnesby and Borgan test.20,21

Risk Stratification and Net BenefitData of the TNT and IDEAL trials were combined to determine the ability of the derived model to stratify risk and treatment effect. Predictions of 5-year MCVE risk while on usual-dose statin treat-ment were made for all trial participants and displayed in a histogram. Similarly, the distribution of predictions of 5-year effect of high-dose versus usual-dose statin treatment was presented in a histogram.

Moreover, we determined the incremental value of treatment decision-making on the basis of predicted treatment effect. Such incremental value is conditional on the estimated harm of unnecessary high-dose statin treatment (ie, adverse events, inconvenience, and costs) and can be expressed in terms of net benefit.22 By unnecessary treatment we mean high-dose statin treatment of patients who will not have a MCVE even on usual-dose statin treatment. When unnecessary treatment is assumed to be relatively harmless, there is no need for selection and, thus, no rationale for using a prediction model. However, when unnecessary treatment is considered detrimental, the treatment effect needs to be adjusted for treatment harm (ie, net benefit). To enable adjustment of treatment effect for treatment harm, a quantitative estimate of the latter is needed. For this purpose,


Dow

nloaded from


Dorresteijn et al High-Dose Statins in Individual Patients 2487

treatment harm can be quantified by estimating at which level of MCVE risk the benefits of high-dose statin treatment start to outweigh treatment harm (ie, the cut-off value for determining high risk). For example, when ≥15% MCVE risk is thought to outweigh treatment harm, the threshold is 15%. This threshold reflects the relative weight of unnecessary treatment of low-risk patients versus correct treatment of high-risk patients and is incorporated in the net benefit calculation.22

In this study, we compared the net benefit of selecting patients for high-dose statin therapy based on our prediction model versus the net benefit of treating all patients, no patients or just the patients with the metabolic syndrome or diabetes mellitus with high-dose statin therapy. Metabolic syndrome was defined by the same criteria used by Deedwania et al10 as the presence of ≥3 of the following risk factors: body mass index of ≥28 kg/m2, triglycerides ≥150 mg/dL (1.7 mmol/L), high-density lipoprotein cholesterol ≤40 mg/dL (1.0 mmol/L) in men or ≤50 mg/dL (1.3 mmol/L) in women, blood pres-sure ≥130/85 mm Hg, or fasting glucose ≥100 mg/dL (5.6 mmol/L). We did not evaluate the net benefit of selective administration of high-dose statin treatment to patients whose LDL-cholesterol is not on tar-get (<70 mg/dL or <1.8 mmol/L) during usual-dose statin therapy, as is currently recommended by guidelines.7,8 Because the LDL-cholesterol was <70 mg/dL (≈1.8 mmol/L) in only 4% of trial patients, the net benefit of that strategy is approximately equal to the net benefit of high-dose statin treatment for all patients. Calculations were based on pooled data of control group patients (treated with usual-dose statin) of the TNT and the IDEAL trials. We refrained from making subjec-tive assumptions about the appropriate decision threshold level and present the net benefit for a range of thresholds in a decision curve.

Analyses were conducted with R statistical software version 2.11.1. (http://www.R-project.org) using add-on packages Design, Hmisc, Survival, Survcomp, stdca (http://www.decisioncurveanalyis.org).

ResultsTable 1 shows baseline characteristics of participants of the TNT and IDEAL trials. Patients had a median age of 62 years and were 81% men in both trials. Almost half of patients ful-filled the criteria of the metabolic syndrome. In TNT, baseline lipids were recorded during the run-in phase on atorvastatin 10 mg, whereas in IDEAL baseline lipids were measured with the majority of patients still using lipid-lowering medi-cation (doses not exceeding the equivalent of simvastatin 20 mg) that was prescribed before enrollment. Almost all par-ticipants received aspirin or anticoagulants, and the majority were using antihypertensive medication. In TNT, 982 MCVEs occurred during 47 369 person-years follow-up (yearly event rate: 2.1%). In IDEAL 1141 MCVEs occurred during 39 099 person-years follow-up (yearly event rate: 2.9%).

Model DerivationInteraction terms for sex-based differences and terms for his-tory of abdominal aortic aneurysm and history of percutaneous coronary intervention were removed from the starting model during backward selection. Next, coefficients were penalized by a uniform shrinkage factor of 0.965, and the proportional hazards assumption was confirmed. The 16 coefficients of the final model, accompanying likelihood ratio statistics, proba-bility values, and (unpenalized) hazard ratios with correspond-ing 95% confidence intervals are presented in Table 2. The computational formulas for 5-year treatment effect of high-dose versus usual-dose statin treatment are presented in Table I in the online-only Data Supplement. Moreover, a calculation sheet is available in the online-only Data Supplement, and an example of treatment effect prediction for an individual patient with this calculation sheet is demonstrated in Figure 1.

Model ValidationThe model C statistic was 0.67 (95% CI, 0.65–0.68) in its deri-vation sample (ie, TNT), and 0.63 (95% CI, 0.62–0.65) in the validation sample (ie, IDEAL). The calibration plots of 5-year predicted versus observed event-free survival (ie, 1 risk) in the derivation and validation sample (Figure 2A and 2B) show that model calibration was excellent. The probability values of the Gronnesby and Borgan tests were 0.65 in the deriva-tion sample and 0.30 in the validation sample, thus confirming satisfactory goodness-of-fit.

Table 1. Characteristics of TNT and IDEAL Participants

Characteristic TNT (n=10 001) IDEAL (n=8888) Total (n=18 889)

Age, y 62 [55–68] 62 [55–69] 62 [55–69]

Male sex 8099 (81%) 7187 (81%) 15 286 (81%)

Current smoking 1341 (13%) 1838 (21%) 3179 (17%)

Past smoking 6322 (63%) 5191 (58%) 11 513 (61%)

Systolic blood pressure, mm Hg

130 [120–140] 135 [120–150] 130 [120–145]

Diastolic blood pressure, mm Hg

80 [70–84] 80 [74–87] 80 [70–85]

Body mass index, kg/m2 27.9 [25.5–30.7] 26.9 [24.7–29.4] 27.4 [25.1–30.1]

Medical history

Myocardial infarction

5834 (58%) 8879 (100%) 14 713 (78%)

Cerebrovascular event

517 (5%) 655 (7%) 1172 (6%)

Abdominal aortic aneurysm

101 (1%) 76 (1%) 177 (1%)

Percutaneous coronary intervention

5407 (54%) 2052 (23%) 7459 (40%)

Coronary artery bypass grafting

4654 (47%) 1769 (20%) 6423 (34%)

Congestive heart failure

781 (8%) 537 (6%) 1318 (7%)

Diabetes mellitus 1501 (15%) 1069 (12%) 2570 (14%)

Metabolic syndrome 4686 (47%) 3948 (44%) 8634 (46%)

Laboratory assessment

Total cholesterol, mg/dL

173 [158–190] 193 [170–220] 180 [162–201]

HDL-cholesterol, mg/dL

45 [39–53] 46 [39–54] 45 [39–53]

Triglycerides, mg/dL 135 [101–183] 133 [97–177] 133 [98–181]

eGFR, mL/min/1.73 m2 65.1 [58.1–72.2] 67.8 [59.6–76.4] 66.1 [58.5–74.1]

Medication use

Any antihypertensive medication

7916 (79%) 8006 (90%) 15 922 (84%)

Aspirin 8653 (87%) 7033 (79%) 15 686 (83%)

Statin, before enrolment

6203 (62%) 6713 (76%) 12 916 (68%)

Data are presented as median [interquartile range] or number (percentage). eGFR indicates glomerular filtration rate (estimated by the Modification of Diet in Renal Disease [MDRD] equation); HDL, high-density lipoprotein; IDEAL, Incremental Decrease in End Points Through Aggressive Lipid Lowering trial; and TNT, Treating to New Targets trial.


Dow

nloaded from

http://www.R-project.org

http://www.decisioncurveanalyis.org



Risk StratificationThe median 5-year MCVE risk of patients on usual-dose statin therapy was 11% (interquartile range, 8% to 15%). Figure 3A shows that 44% of patients had ≤10% predicted baseline MCVE risk, 43% had intermediate (10 to <20%) predicted MCVE risk, and 13% had ≥20% predicted 5-year MCVE risk. As a consequence, predicted treatment effect of high-dose versus usual-dose statin therapy also var-ied widely (Figure 3B). The median absolute reduction of 5-year MCVE risk was 2.2% (interquartile range, 1.6–3.1%; 5-year number needed to treat [NNT] 45). However, 41.9% of patients had <2% predicted treatment effect (5-year NNT>50) and 46.4% of patients had 2 to 4% predicted treat-ment effect (5-year NNT between 25 and 50). A total of 11.7% of patients had >4% predicted reduction of absolute 5-year MCVE risk (5-year NNT <25).

The decision curve in Figure 4 presents the net benefit of decision-making on the basis of model predictions at a range of thresholds of clinical relevance. The appropriate decision-threshold level is subjective and conditional on the estimated harm (ie, adverse events, inconvenience, and costs) of unnec-essary high-dose statin treatment. A low threshold is appro-priate when estimated harms are low and vice versa. For threshold levels between ≈6% and ≈23% 5-year MCVE-risk, prediction-based treatment is associated with higher net ben-efit than making the same decisions for all patients (ie, ator-vastatin 80 mg for all or usual-dose statin treatment for all).

In other words, Figure 4 shows that if treatment of patients with <6% 5-year MCVE-risk with atorvastatin 80 mg instead of usual-dose statin is acceptable, the strategy that leads to optimal net benefit is to simply prescribe atorvastatin 80 mg to all. The median treatment effect will then be 2.0% reduction of absolute MCVE risk (effective NNT=49; Table 3). If a threshold between ≈6% and ≈23% is considered appropriate, more selective prediction-based treatment can reduce the treatment rate and average NNT. For example, if treatment is reserved for patients with ≥15% 5-year MCVE risk, the treatment rate is only 26% of CAD patients and the effective NNT decreases to 26 (Table 3). If a 5-year NNT of ≥21 is still considered unacceptably high, our prediction model is not able to identify patients whose MCVE risk is high enough for the benefits of treatment to outweigh treatment harm. In that situation, maximum net benefit is achieved by treating all patients with usual-dose statins. Selective high-dose treatment of patients with the metabolic syndrome or diabetes mellitus is not associated with the highest net benefit at any decision-threshold level. If this were the strategy of choice, however, the treatment rate would be 49% and the average treatment effect 2.8% (NNT=36).

DiscussionThe present study shows that prediction of the incremental treatment effect of high-dose versus usual-dose statin therapy in individual coronary artery disease patients by means of a multivariable prediction model enables selection of high-risk patients that benefit to a larger extent than other participants in the study. The prediction model presented in this article is based on 13 clinical patient characteristics that are readily available in clinical practice without the need for additional diagnostic tests. Based on this prediction model we show that the 5-year MCVE risk varies widely and, thus, that the predicted treatment effect of high-dose statin therapy in terms of absolute reduction of 5-year MCVE risk has a wide range, even within the seemingly homogeneous group of patients with stable CAD who participated in the TNT and IDEAL trials. Predicted treatment effect was <2% reduction of 5-year MCVE risk in 41.9% of those patients, meaning that their NNT is 50 or higher. In contrast, a group of 11.7% of patients having >4% predicted treatment effect was identified, indicating that their NNT is ≤25. In addition, we demonstrate that the discrimination and calibration of our prediction model remain adequate when applied in an external population. Our prediction model can therefore be used in clinical practice. Finally, we show that the incremental value of prediction-based treatment is conditional on the estimated

Table 2. Coefficients of the Treatment Effect Prediction Model

Model Coefficient*

LRT Statistic P Value

Hazard Ratio (95% CI)*

Age, y −0.0478 12.8† <0.01† 1.24 [1.10–1.40]†

Age squared, y 0.0005

Male sex 0.3147 12.8 <0.01 1.39 [1.15–1.66]

History of myocardial infarction

0.4097 38.2 <0.01 1.53 [1.33–1.75]

History of CABG 0.2263 12.1 <0.01 1.26 [1.11–1.44]

History of congestive heart failure

0.4694 25.3 <0.01 1.62 [1.36–1.95]

History of cerebrovascular disease

0.6172 32.4 <0.01 1.89 [1.54–2.32]

Diabetes mellitus 0.4318 30.8 <0.01 1.56 [1.34–1.82]

Current smoking 0.5379 38.2 <0.01 1.75 [1.48–2.07]

Total cholesterol, per 10 mg/dL

0.0042 9.5 <0.01 1.04 [1.02–1.07]

HDL-cholesterol, per 10 mg/dL

−0.0130 15.0 <0.01 0.87 [0.82–0.94]

eGFR, mL/min/1.73m2 −0.0605 11.7* <0.01* 0.92 [0.84–1.00]*

eGFR squared, mL/ min/1.73m2

0.0004

Systolic blood pressure (treated or not), per 10 mm Hg

0.0371 3.3 0.07 1.04 [1.00–1.08]

Systolic blood pressure (if on treatment), per 10 mm Hg

0.0254 14.5 <0.01 1.03 [1.01–1.04]

High-dose statin, vs usual dose statin

−0.2411 15.2 <0.01 0.78 [0.69–0.88]

CABG indicates coronary artery bypass grafting; CI, confidence interval; eGFR, glomerular filtration rate (estimated by the Modification of Diet in Renal Disease [MDRD] equation; LRT, Likelihood Ratio Test; and HDL, high-density lipoprotein.

*Uniform shrinkage was applied to the coefficients but not the hazard ratios, because penalization increases external validity of the model overall, yet leads to underestimation of the importance of the individual risk factors.

†The risk model contains linear and squared terms for age and eGFR. Therefore, the χ2 and P value of age and eGFR are computed for both terms together. Furthermore, hazard ratios were computed for the difference between the study population’s 75th and the 25th percentile of age (ie, 69 versus 55 yr) and eGFR (ie, 72 versus 58 mL/min/1.73 m).2


Dow

nloaded from



harm that is associated with high-dose statin treatment. The estimated harm of treatment needs to be quantified in terms of a decision-threshold. If the threshold ranges between ≈6% and ≈23%, selective treatment with high-dose statins of high-risk patients (ie, whose predicted treatment effect is higher than average) results in higher net benefit than using the same statin dose (high or low) for everyone. Moreover, prediction-based treatment led to higher net benefit than selective treatment based on the presence of the metabolic syndrome or diabetes mellitus,10 meaning that the present risk score provides a superior estimate of absolute risk and treatment effect. Individual treatment-effect prediction can thus be used to reduce the treatment rate, reduce the effective NNT, and increase the average treatment effect of patients who receive high-dose statins such as atorvastatin 80 mg.

This post hoc analysis of the TNT and IDEAL trials addresses an ongoing discussion about appropriate indications for high-dose statin treatment in patients with stable CAD. Historically, results of randomized controlled trials are imple-mented in clinical practice by either applying the intervention to all patients (in the case of a positive trial result) or to no one (in the case of a negative result). The situation of high-dose statin treatment in patients with stable CAD, however, is different. TNT, IDEAL, and a wealth of other studies have provided consistent high-quality evidence of superior effec-tiveness of high-dose statin over usual-dose statin therapy.6 Moreover, there are no indications that effectiveness is lack-ing in subgroups of CAD patients,6 except perhaps those with severe congestive heart failure (NYHA class II–IV),23,24 aortic stenosis,25 or those undergoing hemodialysis.26 Still, clinicians

Age 65 Years

Sex

Current Smoking

Systolic Blood Pressure 130 mmHg

On antihypertensive treatment?

Medical History:

Diabetes Mellitus 5-yr NNT* =25Hx of myocardial infarction

Hx of CABG

Hx of congestive heart failure

Hx of cerebrovascular disease

Laboratory Results:

HDL-cholesterol mmol/L or 40 mg/dL

Total cholesterol mmol/L or 230 mg/dL

eGFR 50 mL/min

*Atorvastatin 80mg versus usual-dose statin

14%

4%

0%

10%

20%

30%

40%

5-yr Tx effect*

Residual 5-yr MCVE-risk(not preventable byatorvastatin 80mg)

5-year MCVE-risk on usual-dose statin(will appear once all characteristics are completed)

Male

No

No

Yes

Yes

No

No

Yes

Figure 1. Example of high-dose statin treatment effect prediction for an individual patient with the calculation sheet available online. This 65-year-old nonsmoking male with a medical history of coronary artery bypass grafting (CABG) after myocardial infarction who has on-target blood pressure while on medication, stage 3 kidney disease, and whose cholesterol is not on-target despite low-dose statin therapy, has 18% 5-year major cardiovascular event (MCVE) risk. High-dose statin treatment could reduce his MCVE risk by 4%, meaning that the number of similar patients that need to be treated during 5 years to prevent one MCVE (NNT) is 25. This patient, however, will still have 14% residual MCVE risk. eGFR indicates glomerular filtration rate; HDL, high-density lipoprotein; and Tx, treatment.

0.70 0.75 0.80 0.85 0.90 0.95 1.00

0.70

0.75

0.80

0.85

0.90

0.95

1.00

Predicted 5−year event free survival

Obs

erve

d 5−

year

eve

nt fr

ee s

urvi

val

�

�

�

�

� �

�

�

��

0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00

0.65

0.70

0.75

0.80

0.85

0.90

0.95

1.00

Predicted 5−year event free survival

Obs

erve

d 5−

year

eve

nt fr

ee s

urvi

val

�

�

�

�

��

� � �

�

A B

Figure 2. Calibration plots. Predicted and observed event-free survival (ie, 1 minus risk) within deciles of predicted survival at 5-years in the derivation sample (ie, the Treating to New Targets trial [TNT] cohort; A) and the validation sample (ie, the Incremental Decrease in End Points Through Aggressive Lipid Lowering trial [IDEAL] cohort; B).


Dow

nloaded from



are reluctant to use high-dose statin treatment and only do it in a selective subset of CAD-patients, most often those not reach-ing treatment targets with usual-dose statins. Reasons include a higher frequency of side effects, although the magnitude of this association is not exactly known. Higher rates of myopathy were observed during simvastatin 80 mg compared with sim-vastatin 20 mg in the Study of the Effectiveness of Additional Reductions in Cholesterol and Homocysteine (SEARCH) trial, but it is often argued that this problem could be specific to sim-vastatin.3 Moreover, higher rates of myalgia, diarrhea, abdomi-nal pain, and nausea were observed during atorvastatin 80 mg in the IDEAL trial, but this was an open-label trial which could have influenced adverse event reporting.2 Still, higher rates of adverse reactions were also observed during double-blind treatment with atorvastatin 80 mg versus atorvastatin 10 mg in the TNT trial (8.1% versus 5.8%, P<0.001), and this led to a higher discontinuation rate.1 In addition, high-dose statins are associated with a 12% increased relative risk of new-onset dia-betes mellitus.9 A meta-analysis, however, demonstrates that this translates to only a small absolute risk increase: the aver-age 1-year number needed to harm was 498.9

Nevertheless, treatment guidelines endorse the clinicians’ reluctance and recommend reserving high-dose statin treatment for patients whose LDL-cholesterol is not on target during usual-dose statin treatment.7,8 Yet, because cholesterol treatment is aimed at reducing MCVEs, not just lowering cholesterol, this apparently obvious recommendation may not be entirely logical. In fact, meta-analyses demonstrate that the relative treatment effect of statin therapy for reducing MCVEs is not dependent on baseline LDL-cholesterol level.6 Importantly, in the absence of such modification of relative treatment effect, absolute treatment effect of high-dose statin therapy (ie, reduction of absolute MCVE risk) is proportional to the patient’s absolute MCVE risk. Of course, LDL-cholesterol level can be considered as one of the indicators of risk in this respect. However, other patient characteristics or multiple

characteristics combined may be superior predictors of MCVE risk.

Barriers toward widespread implementation of prediction models in clinical practice include the complexity of com-puting treatment effect for individual patients and the fact that it is time-consuming. Simple decision-algorithms are, therefore, preferred, whereas 13 patient-specific variables were required for treatment effect estimation using the model

0 5 10 15 20 25 30

0.00

0.02

0.04

0.06

0.08

0.10

0.12

Decision−threshold (% 5−yr MCVE−risk on usual−dose statin)

Net

ben

efit

All usual−dose statin TxAll atorvastatin 80mgSelective high−dose Tx of MS/DM−patientsPrediction−based Tx

Figure 4. Decision curve. Graphical representation of net benefit at a range of decision thresholds. Net benefit is a measure by which the correct identification of high-risk patients is weighed against over-diagnosis of nondisease. The relative weight of over-diagnosis of nondisease (leading to unnecessary treatment) is dependent on the severity of treatment harm (ie, adverse events, inconvenience, and costs). These harms are reflected by the decision threshold. For example, when treatment harm is thought to be equivalent to the treatment effect in patients having 15% major cardiovascular event (MCVE) risk, the decision-threshold is 15%. DM indicates diabetes mellitus; LDLc, low-density lipoprotein cholesterol; MS, metabolic syndrome; and Tx, treatment.

5−year MCVE−risk on usual−dose statin

Freq

uenc

y

0 to <10% risk: 44.1% of patients10 to <20% risk: 43.3% of patients20 to <30% risk: 9.3% of patients30 to <40% risk: 2.4% of patients40% risk or higher: 0.9% of patients

0%

A B

10% 20% 30% 40% 50%

010

0020

0030

0040

00

Treatment effect of high versus usual−dose statin (ARR for MCVE)

Freq

uenc

y

NNT >50: 41.9% of patientsNNT 25 to 50: 46.4% of patientsNNT 17 to 25: 9.4% of patientsNNT 12 to 17: 2.0% of patientsNNT 12 or lower: 0.3% of patients

0% 2% 4% 6% 8% 10%

010

0020

0030

0040

00

Figure 3. Distribution of predicted baseline major cardiovascular event (MCVE) risk and treatment effect of high-dose versus usual-dose statin that is derived thereof. A, Distribution of predicted baseline 5-year MCVE risk among all participants of the Treating to New Targets trial (TNT) and Incremental Decrease in End Points Through Aggressive Lipid Lowering trial (IDEAL) trials (given usual-dose statin use). B, Deducted from A: distribution of predicted 5-year treatment effect of atorvastatin 80 mg (5-year absolute risk reduction [ARR] for MCVEs). NNT indicates number needed to treat.


Dow

nloaded from



presented in this study. However, the patient characteristics that are required for estimation of treatment effect with the present prediction model are easy to measure and usually already available in practice. Furthermore, the calculation sheet presented in the online-only Data Supplement facilitates treatment effect estimation in the consulting room by simply entering these patient specific variables.

Apart from improvement of net benefit, individualized treatment effect prediction may have important additional advantages. First, it can be used by the clinician to educate patients about their individual prognosis and risk factors for recurrent events. This may improve patients’ understanding of the disease and the need for preventive treatment. Second, it enables patients to be involved in treatment decision-making. The calculation sheet’s interactive bar graph demonstrates potential treatment effect in an insightful manner. Because patients’ disagreement with the need for treatment is one of the main reasons for nonadherence,27 shared decision-making may increase patient compliance. This is important because the adherence-rate to self-administered treatments is typically only 50%, thereby seriously compromising the effect of preventive therapy.27 Experience with a similar health decision aid shows that implementation in routine clinical practice is feasible, increases patients’ perception of shared decision-making, and reduces patients’ anxiety.28,29

Limitations of the present study include the fact that our prediction model is derived from a trial population that was selected on the basis of strict eligibility criteria. Participants of the TNT trial were, for example, aged ≤75 years, were not allowed to have a left ventricular ejection fraction <30%, and did not have any survival-limiting diseases other than CAD.11 It is, therefore, uncertain whether our prediction model can be generalized to patients who do not fit the TNT trial eligibility profile. Furthermore, because atherosclerosis is a systemic progressive condition, many CAD patients have multiple manifestations of disease. Although concomitant cerebrovascular disease, aortic aneurysmal disease, and peripheral artery disease were allowed, the prevalence of these conditions in the TNT and IDEAL populations was low. For example, an abdominal aortic aneurysm was present in only 101 TNT participants (1%). Although we know from previous research that having multiple disease manifestations is a strong

risk factor for recurrent vascular events,30 the low prevalence of abdominal aortic aneurysm in this population explains why its regression coefficient did not reach statistical significance and was removed from the model. Participants of the IDEAL trial were even more homogeneous, because all had a history of a definite myocardial infarction.12 This explains why the C statistic of our prediction model was considerably lower in the IDEAL dataset than in the derivation set. In fact, the C statistics that we found in this study may be considered low for a risk score in an asymptomatic population. Yet, the distribution of predicted treatment effect (Figure 3B) shows that the model can still discriminate patients whose expected gain from high-dose statin treatment is considerably higher than average. Furthermore, the decision curve (Figure 4) demonstrates that using model predictions for treatment assignment can improve net benefit at clinically relevant threshold levels. We recognize that the model’s discrimination could be further improved by including additional information on biomarkers, such as C-reactive protein (hs-CRP),31 or imaging, such as carotid intima-media thickness or coronary CT-angiography.32,33 Yet, because such information is not readily available for most patients, this would come at the expense of more limited applicability in clinical practice. Also, because of the limited duration of follow-up in trials, models like these can only predict treatment effect for a certain duration of follow-up (in this case 5 years), whereas statins are usually prescribed lifelong in practice. It may be argued that patients with low predicted 5-year treatment effect still benefit from high-dose statins after a longer time period. Finally, although the treatment arms of the trial were pooled for derivation of the prediction model, this does not violate the randomized distribution of the study interventions. Therefore, we stress that treatment effect predictions on the basis of this model are not vulnerable to confounding bias.

ConclusionsThe incremental treatment effect of high-dose statin therapy over usual-dose statin therapy in individual CAD patients can be estimated by a prediction model, containing 13 easy-to-measure clinical predictors that are readily available in clini-cal practice. Predicted treatment effect can be used to guide treatment decisions in clinical practice, and this may improve net benefit of therapy.

Sources of FundingThis study was not supported by any funding. The TNT and IDEAL trials were funded by Pfizer. The sponsors of the contributing trials provided the requested data, but they did not play any role in the statistical analysis.

DisclosuresDr Boekholdt reports receipt of consultancy fees from Pfizer. Dr Kastelein reports receipt of lecture honoraria from Merck Sharpe & Dohme, Roche, Novartis, ISIS, Genzyme, Pfizer, Kowa, and AstraZeneca. Dr LaRosa reports receipt of consultancy fees from Pfizer and Amgen, and travel expenses from Pfizer. Dr Pedersen reports receipt of research grant support and lecture fees from Pfizer. Dr DeMicco reports having stock/stock options with and being a full-time employee of Pfizer. Dr Ridker reports receipt of research grant funding from Novartis and AstraZeneca; serving as a consul-tant to ISIS, Vascular Biogenics, Merck Sharpe & Dohme, Abbott,

Table 3. Inferences and Consequences for Clinical Practice

Tx Threshold (5-Year MCVE Risk)

Tx Strategy AssociatedWith Optimal NB Tx Rate*

Median Tx Effect(ARR)†

5-Year NNT to Prevent 1

MCVE†

<6% Treat all with high-dose statin

100% 2.0% 49

10% Prediction-based Tx 56% 2.9% 34



>23% Treat all with usual-dose statin

0% NA NA

AAR indicates absolute risk reduction; MCVE, major cardiovascular event; NB, net benefit; NNT, number needed to treat; and Tx, treatment.

*Percentage treated with atorvastatin 80 mg instead of usual-dose statin. †Median ARR for MCVE achieved by atorvastatin 80 mg vs usual-dose statin

in treated patients


Dow

nloaded from



and Boerringer-Ingelheim; board membership with Merck Sharpe & Dohme; receipt of a grant or pending grant to his institution from Amgen; and being listed as a coinventor on patents held by the Brigham and Women’s Hospital that relate to the use of inflamma-tory biomarkers in cardiovascular disease and diabetes mellitus that have been licensed to AstraZeneca and Siemens. Dr Visseren’s de-partment receives grant support from The Netherlands Organisation for Health Research and Development and the Catharijne founda-tion Utrecht, and speaker fees from Merck. The other authors report no conflicts.

References 1. LaRosa JC, Grundy SM, Waters DD, Shear C, Barter P, Fruchart JC, Gotto

AM, Greten H, Kastelein JJ, Shepherd J, Wenger NK; Treating to New Targets (TNT) Investigators. Intensive lipid lowering with atorvastatin in patients with stable coronary disease. N Engl J Med. 2005;352:1425–1435.

2. Pedersen TR, Faergeman O, Kastelein JJ, Olsson AG, Tikkanen MJ, Holme I, Larsen ML, Bendiksen FS, Lindahl C, Szarek M, Tsai J; Incremental Decrease in End Points Through Aggressive Lipid Lowering (IDEAL) Study Group. High-dose atorvastatin vs usual-dose simvastatin for secondary prevention after myocardial infarction: the IDEAL study: a randomized controlled trial. JAMA. 2005;294:2437–2445.

3. Armitage J, Bowman L, Wallendszus K, Bulbulia R, Rahimi K, Haynes R, Parish S, Peto R, Collins R. Intensive lowering of LDL cholesterol with 80 mg versus 20 mg simvastatin daily in 12,064 survivors of myocardial infarction: a double-blind randomised trial. Lancet. 2010;376:1658–1669.

4. Cannon CP, Braunwald E, McCabe CH, Rader DJ, Rouleau JL, Belder R, Joyal SV, Hill KA, Pfeffer MA, Skene AM; Pravastatin or Atorvastatin Evaluation and Infection Therapy-Thrombolysis in Myocardial Infarction 22 Investigators. Intensive versus moderate lipid lowering with statins after acute coronary syndromes. N Engl J Med. 2004;350:1495–1504.

5. de Lemos JA, Blazing MA, Wiviott SD, Lewis EF, Fox KA, White HD, Rouleau JL, Pedersen TR, Gardner LH, Mukherjee R, Ramsey KE, Palmisano J, Bilheimer DW, Pfeffer MA, Califf RM, Braunwald E; Inves-tigators. Early intensive vs a delayed conservative simvastatin strategy in patients with acute coronary syndromes: phase Z of the A to Z trial. JAMA. 2004;292:1307–1316.

6. Baigent C, Blackwell L, Emberson J, Holland LE, Reith C, Bhala N, Peto R, Barnes EH, Keech A, Simes J, Collins R. Efficacy and safety of more in-tensive lowering of LDL cholesterol: a meta-analysis of data from 170,000 participants in 26 randomised trials. Lancet. 2010;376:1670–1681.

7. Smith SC Jr, Allen J, Blair SN, Bonow RO, Brass LM, Fonarow GC, Grundy SM, Hiratzka L, Jones D, Krumholz HM, Mosca L, Pasternak RC, Pearson T, Pfeffer MA, Taubert KA; AHA/ACC; National Heart, Lung, and Blood Institute. AHA/ACC guidelines for secondary prevention for patients with coronary and other atherosclerotic vascular disease: 2006 update: endorsed by the National Heart, Lung, and Blood Institute. Circu-lation. 2006;113:2363–2372.

8. Perk J, De Backer G, Gohlke H, Graham I, Reiner Z, Verschuren M, Albus C, Benlian P, Boysen G, Cifkova R, Deaton C, Ebrahim S, Fisher M, Germano G, Hobbs R, Hoes A, Karadeniz S, Mezzani A, Prescott E, Ryden L, Scherer M, Syvanne M, Scholte Op Reimer WJ, Vrints C, Wood D, Zamorano JL, Zannad F, Cooney MT, Bax J, Baumgartner H, Ceconi C, Dean V, Fagard R, Funck-Brentano C, Hasdai D, Kirchhof P, Knuuti J, Kolh P, McDonagh T, Moulin C, Popescu BA, Sechtem U, Sirnes PA, Tendera M, Torbicki A, Vahanian A, Windecker S, Aboyans V, Ezquerra EA, Baigent C, Brotons C, Burell G, Ceriello A, De Sutter J, Deckers J, Del Prato S, Diener HC, Fitzsimons D, Fras Z, Hambrecht R, Jankowski P, Keil U, Kirby M, Larsen ML, Mancia G, Manolis AJ, McMurray J, Pajak A, Parkhomenko A, Rallidis L, Rigo F, Rocha E, Ruilope LM, van der Velde E, Vanuzzo D, Viigimaa M, Volpe M, Wiklund O, Wolpert C. European Guidelines on cardiovascular disease prevention in clinical practice (version 2012): The Fifth Joint Task Force of the European Society of Cardiology and Other Societies on Cardiovascular Disease Prevention in Clinical Practice (constituted by representatives of nine societies and by invited experts) * Developed with the special contribution of the European Association for Cardiovascular Prevention & Rehabilitation (EACPR). Eur Heart J. 2012;33:1635–1701.

9. Preiss D, Seshasai SR, Welsh P, Murphy SA, Ho JE, Waters DD, DeMicco DA, Barter P, Cannon CP, Sabatine MS, Braunwald E, Kastelein JJ, de Lemos JA, Blazing MA, Pedersen TR, Tikkanen MJ, Sattar N, Ray KK.

Risk of incident diabetes with intensive-dose compared with moderate-dose statin therapy: a meta-analysis. JAMA. 2011;305:2556–2564.

10. Deedwania P, Barter P, Carmena R, Fruchart JC, Grundy SM, Haffner S, Kastelein JJ, LaRosa JC, Schachner H, Shepherd J, Waters DD; Treating to New Targets Investigators. Reduction of low-density lipoprotein cholesterol in patients with coronary heart disease and metabolic syndrome: analysis of the Treating to New Targets study. Lancet. 2006;368:919–928.

11. Waters DD, Guyton JR, Herrington DM, McGowan MP, Wenger NK, Shear C; TNT Steering Committee Members and Investigators. Treating to New Targets (TNT) Study: does lowering low-density lipoprotein cho-lesterol levels below currently recommended guidelines yield incremental clinical benefit? Am J Cardiol. 2004;93:154–158.

12. Pedersen TR, Faergeman O, Kastelein JJ, Olsson AG, Tikkanen MJ, Holme I, Larsen ML, Bendiksen FS, Lindahl C, Palmer G; Incremental Decrease in End Points Through Aggressive Lipid Lowering Study Group. Design and baseline characteristics of the Incremental Decrease in End Points through Aggressive Lipid Lowering study. Am J Cardiol. 2004;94:720–724.

13. Dorresteijn JA, Visseren FL, Ridker PM, Paynter NP, Wassink AM, Buring JE, van der Graaf Y, Cook NR. Aspirin for primary prevention of vascular events in women: individualized prediction of treatment effects. Eur Heart J. 2011;32:2962–2969.

14. Dorresteijn JA, Visseren FL, Ridker PM, Wassink AM, Paynter NP, Steyerberg EW, van der Graaf Y, Cook NR. Estimating treatment effects for individual patients based on the results of randomised clinical trials. BMJ. 2011;343:d5888.

15. D’Agostino RB Sr, Vasan RS, Pencina MJ, Wolf PA, Cobain M, Massaro JM, Kannel WB. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation. 2008;117:743–753.

16. K/DOQI clinical practice guidelines for chronic kidney disease: evalua-tion, classification, and stratification. Am J Kidney Dis. 2002;39:S1–266.

17. Donders AR, van der Heijden GJ, Stijnen T, Moons KG. Review: a gentle introduction to imputation of missing values. J Clin Epidemiol. 2006;59:1087–1091.

18. Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measur-ing and reducing errors. Stat Med. 1996;15:361–387.

19. Steyerberg EW. Clinical prediction models: a practical approach to devel-opment, validation, and updating. New York: Springer; 2009.

20. Grønnesby JK, Borgan O. A method for checking regression mod-els in survival analysis based on the risk score. Lifetime Data Anal. 1996;2:315–328.

21. May S, Hosmer DW. A simplified method of calculating an overall good-ness-of-fit test for the Cox proportional hazards model. Lifetime Data Anal. 1998;4:109–120.

22. Steyerberg EW, Pencina MJ, Lingsma HF, Kattan MW, Vickers AJ, Van Calster B. Assessing the incremental value of diagnostic and prognostic markers: a review and illustration. Eur J Clin Invest. 2012;42:216–228.

23. Kjekshus J, Apetrei E, Barrios V, Böhm M, Cleland JG, Cornel JH, Dunselman P, Fonseca C, Goudev A, Grande P, Gullestad L, Hjalmarson A, Hradec J, Jánosi A, Kamenský G, Komajda M, Korewicki J, Kuusi T, Mach F, Mareev V, McMurray JJ, Ranjith N, Schaufelberger M, Vanhaecke J, van Veldhuisen DJ, Waagstein F, Wedel H, Wikstrand J; CORONA Group. Rosuvastatin in older patients with systolic heart failure. N Engl J Med. 2007;357:2248–2261.

24. Tavazzi L, Maggioni AP, Marchioli R, Barlera S, Franzosi MG, Latini R, Lucci D, Nicolosi GL, Porcu M, Tognoni G. Effect of rosuvastatin in patients with chronic heart failure (the GISSI-HF trial): a randomised, double-blind, placebo-controlled trial. Lancet. 2008;372:1231–1239.

25. Rossebø AB, Pedersen TR, Boman K, Brudi P, Chambers JB, Egstrup K, Gerdts E, Gohlke-Bärwolf C, Holme I, Kesäniemi YA, Malbecq W, Nienaber CA, Ray S, Skjaerpe T, Wachtell K, Willenheimer R; SEAS Investigators. Intensive lipid lowering with simvastatin and ezetimibe in aortic stenosis. N Engl J Med. 2008;359:1343–1356.

26. Fellström BC, Jardine AG, Schmieder RE, Holdaas H, Bannister K, Beutler J, Chae DW, Chevaile A, Cobbe SM, Grönhagen-Riska C, De Lima JJ, Lins R, Mayer G, McMahon AW, Parving HH, Remuzzi G, Samuelsson O, Sonkodi S, Sci D, Süleymanlar G, Tsakiris D, Tesar V, Todorov V, Wiecek A, Wüthrich RP, Gottlow M, Johnsson E, Zannad F; AURORA Study Group. Rosuvastatin and cardiovascular events in pa-tients undergoing hemodialysis. N Engl J Med. 2009;360:1395–1407.

27. Haynes RB, Ackloo E, Sahota N, McDonald HP, Yao X. Interventions for enhancing medication adherence. Cochrane Database Syst Rev. 2008:CD000011.


Dow

nloaded from



28. Decker C, Arnold SV, Olabiyi O, Ahmad H, Gialde E, Luark J, Riggs L, DeJaynes T, Soto GE, Spertus JA. Implementing an innovative consent form: the PREDICT experience. Implement Sci. 2008;3:58.

29. Arnold SV, Decker C, Ahmad H, Olabiyi O, Mundluru S, Reid KJ, Soto GE, Gansert S, Spertus JA. Converting the informed consent from a per-functory process to an evidence-based foundation for patient decision making. Circ Cardiovasc Qual Outcomes. 2008;1:21–28.

30. Steg PG, Bhatt DL, Wilson PW, D’Agostino R Sr, Ohman EM, Röther J, Liau CS, Hirsch AT, Mas JL, Ikeda Y, Pencina MJ, Goto S; REACH Reg-istry Investigators. One-year cardiovascular event rates in outpatients with atherothrombosis. JAMA. 2007;297:1197–1206.

31. Ray KK, Cannon CP, Cairns R, Morrow DA, Ridker PM, Braunwald E. Prognostic utility of apoB/AI, total cholesterol/HDL, non-HDL choles-terol, or hs-CRP as predictors of clinical risk in patients receiving statin therapy after acute coronary syndromes: results from PROVE IT-TIMI 22. Arterioscler Thromb Vasc Biol. 2009;29:424–430.

32. Wattanakit K, Folsom AR, Chambless LE, Nieto FJ. Risk factors for car-diovascular event recurrence in the Atherosclerosis Risk in Communities (ARIC) study. Am Heart J. 2005;149:606–612.

33. Perrone-Filardi P, Musella F, Savarese G, Cecere M, Marciano C, Scala O, Rengo G, Dellegrottaglie S, Cuocolo A, Leosco D. Coronary computed tomography: current role and future perspectives for cardiovascular risk stratification. Eur Heart J Cardiovasc Imaging. 2012;13:453–458.

CLINICAL PERSPECTIVEEvidence from large clinical trials supports the use of high-dose statins in patients with stable coronary artery disease. Still, because of concern about high-dose statins causing higher rates of side effects, such as myalgia, elevated transaminases, and diabetes mellitus, high-dose statins are not unequivocally recommended for all coronary artery disease patients. This study describes the development and validation of a prediction model for the incremental treatment effect of high-dose statins for individual patients in terms of 5-year absolute risk reduction for myocardial infarction, stroke, coronary death, or cardiac resuscitation. Model derivation was based on data from the Treating to New Targets trial (TNT; n=10 001), and the model was validated in the Incremental Decrease in End Points Through Aggressive Lipid Lowering trial (IDEAL; n=8888). The model is easy to use in clinical practice because it contains only readily available predictors, such as age, sex, smoking sta-tus, blood pressure, medical history, and routine laboratory tests. Moreover, it is made available as an interactive calculation sheet. The model can thus be used in the consulting room. A decision curve shows that making treatment decisions on the basis of the model’s predictions may improve the net benefit of treatment, meaning that prediction-based treatment reduces treatment rate and the effective number needed to treat. Estimation of the incremental treatment effect of high-dose versus usual-dose statin therapy in individual coronary artery disease patients thus enables selection of high-risk patients that ben-efit most from more aggressive lipid-lowering therapy.


Dow

nloaded from


Frank L.J. VisserenJohn C. LaRosa, Terje R. Pedersen, David A. DeMicco, Paul M Ridker, Nancy R. Cook and

Johannes A.N. Dorresteijn, S. Matthijs Boekholdt, Yolanda van der Graaf, John J.P. Kastelein,Right Patients Based on Individualized Prediction of Treatment Effect

High-Dose Statin Therapy in Patients With Stable Coronary Artery Disease: Treating the

Print ISSN: 0009-7322. Online ISSN: 1524-4539 Copyright © 2013 American Heart Association, Inc. All rights reserved.

is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231Circulation doi: 10.1161/CIRCULATIONAHA.112.000712

2013;127:2485-2493; originally published online May 14, 2013;Circulation.

http://circ.ahajournals.org/content/127/25/2485World Wide Web at:

The online version of this article, along with updated information and services, is located on the

http://circ.ahajournals.org/content/suppl/2013/05/14/CIRCULATIONAHA.112.000712.DC1Data Supplement (unedited) at:

http://circ.ahajournals.org//subscriptions/

is online at: Circulation Information about subscribing to Subscriptions:

http://www.lww.com/reprints Information about reprints can be found online at: Reprints:

document. Permissions and Rights Question and Answer this process is available in the

click Request Permissions in the middle column of the Web page under Services. Further information aboutOffice. Once the online version of the published article for which permission is being requested is located,

can be obtained via RightsLink, a service of the Copyright Clearance Center, not the EditorialCirculationin Requests for permissions to reproduce figures, tables, or portions of articles originally publishedPermissions:


Dow

nloaded from

http://circ.ahajournals.org/content/127/25/2485

http://circ.ahajournals.org/content/suppl/2013/05/14/CIRCULATIONAHA.112.000712.DC1

http://www.ahajournals.org/site/rights/

http://www.lww.com/reprints

http://circ.ahajournals.org//subscriptions/


SUPPLEMENTAL MATERIAL TO

High-dose statin therapy in patients with stable coronary artery disease: treating the

right patients based on individualized prediction of treatment effect.

J.A.N. Dorresteijn, S.M. Boekholdt, Y. van der Graaf, J.J.P. Kastelein, J.C. LaRosa, T.R. Pedersen, D.A. DeMicco, P.M Ridker,

N.R. Cook, F.L.J. Visseren

Supplemental table 1

Computational formula for 5-year absolute treatment effect of high-dose versus usual-dose statin

treatment in patients with stable coronary artery disease.

Predicted 5-year treatment effect of high-dose statin

=

(1 - 0.78) x 5-year MCVE risk on usual-dose statin

5-year MCVE risk on usual-dose statin (%) = (1- 0.914 exp[A +1.5106]

) x 100%, where

A = -0.0478 x age in years + 0.000515 x (age in years)2 + 0.315 [if male] + 0.410 [if history of

myocardial infarction] + 0.226 [if history of CABG] + 0.469 [if history of congestive heart failure] +

0.617 [if history of cerebrovascular disease] + 0.432 [if diabetic] + 0.538 [if current smoker] + 0.00419

x total cholesterol in mg/dL – 0.0130 x HDL-cholesterol in mg/dL - 0.0605 x eGFR in mL/min/1.73m² +

0.000419 x (eGFR in mL/min/1.73m²)2 + 0.00371 x systolic blood pressure in mmHg + 0.00254 x

systolic blood pressure in mmHg [if on antihypertensive treatment]

high-dose statin therapy in patients with stable coronary...

Documents