Statins for Primary Prevention: Insights from JUPITER Trial CRT 2011, Washington DC Sanjay Kaul, MD Division of Cardiology Cedars-Sinai Heart Institute Professor, Cedars-Sinai Medical Center & Geffen School of Medicine at UCLA Los Angeles, California

Sanjay Kaul, MD Consulting Fees Roche Novo Nordisk

Adapted from Grundy SM, et al. Circulation. 2004;110:227-239. NCEP ATP III Update Establishes New LDL-C Goals and Cut Points 160 mg/dL 130 mg/dL 100 mg/dL Initiate TLC 190 mg/dL 130 mg/dL (100-129 mg/dL: consider Rx ) 100 mg/dL (<100 mg/dL: consider drug Rx) Consider Drug Therapy <160 mg/dL <130 mg/dL <130 mg/dL (optimal <100 mg/dL) <100 mg/dL (optional goal: <70 mg/dL) LDL-C Goal Risk Category Lower risk: 0-1 risk factor Moderate risk: 2+ risk factors (<10%) Moderately high risk: 2+ risk factors (10-year risk 10% -20%) High risk: CHD or CHD risk equivalents (10-year risk >20%) Adapted from Grundy SM, et al. Circulation. 2004;110:227-239.

2009 Canadian Cardiovascular Society (CCS) Guidelines for the Diagnosis and Treatment of Dyslipidemia and Prevention of Cardiovascular Disease in the Adult Primary Goal: LDLC High CAD, CVA, PVD Most pts with Diabetes FRS >20% RRS >20% <2 mmol/L or 50% reduction Class I Level A Moderate FRS 10-19% RRS 10-19% LDL >3.5 mmol/L TC/HDLC >5.0 hsCRP >2 in men >50 yr women >60 yr Class IIA Low FRS <10% <5 mmol/L Secondary Targets TC/HDLC <4, non HDLC <3.5 mol/L hsCRP <2 mg/L, TG <1.7 mol/L, ApoB/A <0.8 Genest J et al. Can J Cardiol 2009;25(10):567-579.

Mosca L et al. Circulation 2011 AHA Primary Prevention Guidelines for Women 2011 Update: Indications for Statins CHD risk Indication Primary Goal: LDLC Recommendation High CAD, CVA, PVD Most pts with Diabetes FRS >20% <100mg/dL <70mg/dL Class I, Level A Class I, Level B Class IIa, Level B Moderate Women >60 yr FRS 10-19% hsCRP >2 Class IIb, Level B Low FRS <10% Mosca L et al. Circulation 2011

Statins for Primary Prevention of CVD Source (year published) Brugts et al. (BMJ 2009) Ray et al. (Arch Int Med 2010) Taylor et al. (Cochrane 2011) Inclusion criteria -RCTs with follow-up ≥1 year; ≤20% had CVD -RCTs from January 1970 to May 2009; participants free from CVD -RCTs from 1994-2006 with treatment duration ≥1 year and follow-up ≥6 months; ≤10% had CVD # of trials / number included (n) 10 RCTs / n=70,388 11 RCTs / n=65,229 14 RCTs; 16 trial arms / n=34,272 JUPITER Included Excluded Follow-up duration Mean 4.1 years Mean 3.7 years, ~244,000 pt-years 1-5.3 years ~113,000 pt years Mean age (years) 63 62 57 % Men 66% 65% 65.9% % Diabetics 23% 19% NR

Statins for Primary Prevention of CVD Source (year published) Brugts et al. (BMJ 2009) Ray et al. (Arch Int Med 2010) Taylor et al. (Cochrane 2011) Total # deaths (statin / control group deaths) 3,650 (1725/1925) 2,793 (1447/1346) 794 (362/432) Baseline mortality events per 100 person years 1.7 1.14 1.0 Satin effect on all-cause mortality (95% CI) OR 0.88 (0.81-0.96); I2=27% RR 0.91 (0.83-1.01); I2=23% RR 0.83 (0.73 to 0.95); I2=8% Statin effect on morbidity Major CHD OR 0.70 (0.61-0.81); I2=60% Major Stroke OR 0.81 (0.71-0.93); I2 =24% Non-fatal MI OR 0.56 (0.41-0.76); I2=50% Not reported Total CVD events RR 0.70 (0.61 to 0.79); I2=0.0% Total CHD events RR 0.72 (0.65-0.79); I2=0.0% Total stoke events RR 0.78 (0.65-0.94); I2=19%

Stains Benefit-Risk Balance in CVD Prevention 2011 Cochrane Review (14 RCTs, N=34,272) Statins Benefit Risk 19% reduction in mortality 30% reduction in CHD events 19% reduction in stroke 44% reduction in nonfatal MI 33% reduction in PCI/CABG 9% increase in diabetes No significant increase in short-term risk of - Muscle adverse events - Liver adverse events - Cancer, memory loss Benefits likely outweigh risks in high-risk patients (>20% 10yr-CHD risk) Sparse cost-effectiveness data Caution is warranted in low-risk patients 8 8

FDA Extended Rosuvastatin Label Approved for the reduction in risk of stroke, myocardial infarction, and revascularization procedures among individuals with normal LDL levels and no clinically evident cardiovascular disease, but with an increased risk based on - Age (>50 yrs in men; >60 yrs in women) - Elevated CRP levels (>2mg/L) - Presence of at least one additional ATP III risk factor http://www.accessdata.fda.gov/drugsatfda_docs/label/2010/021366s016lbl.pdf

Primary Composite Endpoint in JUPITER (first events) Rosuvastatin (n=8901) Control (n=8901) Risk ratio (95% CI) CV death 29 37 0.78 (0.48-1.27) Nonfatal MI 21 61 0.34 (0.21-0.56) Nonfatal stroke 30 57 0.53 (0.34-0.82) Unstable angina 15 27 0.56 (0.30-1.04) Revascularization 47 70 0.67 (0.46-0.97) First MCE 142/8901 252/8901 0.56 (0.46-0.69) 0.5 1 1.5 2 Risk ratio Benefit driven by nonfatal MI, stroke and revascularization Label doesn’t allow claim for CV death or unstable angina

<2 ATP III risk factor >2 ATP III risk factor Treatment Effect by hs CRP and FRS Risk Category: JUPITER Subset Rosuvastatin Control HR (95% CI) P int. hsCRP>4.2mg/L 89/4446 128/4551 0.71 (0.54 to 0.93) 0.015 hsCRP<4.2mg/L 53/4454 124/4350 0.42 (0.3 to 0.57) hsCRP>4mg/L 92/4689 133/4788 0.71 (0.54 to 0.92) 0.014 hsCRP<4mg/L 50/4211 119/4113 0.41 (0.3 to 0.57) hsCRP>3mg/L 111/6252 182/6337 0.62 (0.49 to 0.78) 0.141 hsCRP<3mg/L 31/2649 70/2564 0.43 (0.28 to 0.65) <2 ATP III risk factor 33/2199 35/2080 0.91 (0.56 to 1.46) 0.034 >2 ATP III risk factor 109/6702 217/6821 0.51 (0.41 to 0.64) <10% FRS 29/3615 43/3602 0.67 (0.42 to 1.07) 0.945 10-20% FRS 83/4485 171/4516 0.49 (0.38 to 0.64) >20% FRS 29/786 38/772 0.70 (0.43 to 1.14) 0.5 1 1.5 HR FDA Briefing Document, December 15, 2009

Caveats Challenging The Interpretation of JUPITER Trial Validity of study hypothesis Healthy status of subjects Generalizability to clinical practice Reliability of benefit Mortality benefit Benefit-risk assessment Cost-effectiveness and clinical implications This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Statin Effect According to Baseline LDL & CRP AFCAPS/TexCAPS LDL CRP <median >median 1.08 (0.56-2.08) 2.5% vs 2.2% NNH=333 0.58 (0.34-0.98) 2.9% vs 5.1% NNT=48 0.38 (0.21-0.70) 2.0% vs 5.0% NNT=33 0.68 (0.42-1.10) 3.8% vs 5.5% NNT=58 Statin not effective in low LDL, low CRP group! “Thus, a trial with low LDL & low CRP would be infeasible in terms of power and sample size and highly unlikely to show clinical benefit” Ridker et al, NEJM 2001;344:1959-65

Cardiovascular Event Rates by hs CRP JUPITER <4.2mg/L >4.2mg/L >4mg/L <4mg/L >3mg/L <3mg/L hs CRP Placebo Rosuvastatin =29% =58% =29% =59% =38% =57% Cardiovascular events (%) (median) Flat dose response of hsCRP with events in placebo arm Inverse dose response of hsCRP with statin treatment effect Kaul et al, Arch Int. Med 2010

Treatment Effect by FRS Risk Category and hs CRP: JUPITER Subgroup Rosuvastatin N events Placebo HR (95% CI) P interaction FRS <10%, hsCRP <4.2mg/dL 12 20 0.60 (0.29-1.25) 0.7 FRS <10%, hsCRP >4.2mg/dL 17 40 0.73 (0.39-1.39) FRS 10-20%, hsCRP <4.2mg/dL 30 87 0.34 (0.22-0.52) 0.02 FRS 10-20%, hsCRP >4.2mg/dL 53 84 0.65 (0.46-0.92) FRS >20%, hsCRP <4.2mg/dL 10 0.51 (0.23-1.14) 0.3 FRS >20%, hsCRP >4.2mg/dL 19 23 0.89 (0.47-1.68) Kaul et al, Arch Int. Med 2010

Baseline hs CRP, CVD Risk and Rx effect ASCOT-LLA Subset Cases Control OR (95% CI) P. hsCRP<1.74mg/L 131 448 Reference hsCRP 1.74-4.09mg/L 153 417 1.25 (0.93 to 1.68) 0.14 hsCRP>4.09mg/L 168 404 1.35 (1.00 to 1.82) 0.05 Atorvastatin effect Pint. hsCRP<1.74mg/L 0.60 (0.33 to 1.10) 0.54 hsCRP 1.74-4.09mg/L 0.77 (0.41 to 1.46) hsCRP>4.09mg/L 0.95 (0.51 to 1.78) AUC (-CRP) = 0.620 AUC (+CRP) = 0.627 NRI = 0.03 (P = 0.17) 1 2 3 OR Sever P et al, AHA 2010

On Treatement LDL & hs CRP & CVD Risk ASCOT-LLA Subset Cases Control OR (95% CI) P. On Rx LDL LDL>2.1mmol/L 44 126 1.10 (0.93 to 1.68) 0.68 LDL<2.1mmol/L 23 140 0.45 (0.27 to 0.76) 0.003 Pint = 0.004. On Rx hsCRP hsCRP>1.83mg/L 41 137 0.82 (0.53 to 1.28) 0.38 hsCRP<1.83mg/L 32 140 0.70 (0.43 to 1.14) 0.15 Pint = 0.60. 1 2 3 OR Sever P et al, AHA 2010

Caveats Challenging The Interpretation of JUPITER Trial Validity of JUPITER hypothesis questionable - Without a low-LDL, low-CRP arm, not possible to validate the CRP hypothesis - Baseline CRP not a predictor of outcomes - Benefit might be related exclusively to LDL-lowering (ASCOT) This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Caveats Challenging The Interpretation of JUPITER Trial Validity of study hypothesis Healthy status of subjects Generalizability to clinical practice Reliability of benefit Mortality benefit Benefit-risk assessment Cost-effectiveness and clinical implications This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Caveats Challenging The Interpretation of JUPITER Trial Apparently healthy status of subjects claim disputable - Patients with hsCRP <2 mg/dL (normal) excluded - 41% with metabolic syndrome - 50% overweight (average female 165 lbs; average male 192 lbs) - 25% hypertensive by JNC-VII criterion (140/85 mmHg) - Family history of CVD (11%) - Smokers (16%) - HbA1c not in “normal” range (threshold for CVD risk = 4.7%) - Aspirin use (16.6%) This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Caveats Challenging The Interpretation of JUPITER Trial Validity of study hypothesis Healthy status of subjects Generalizability to clinical practice Reliability of benefit Mortality benefit Benefit-risk assessment Cost-effectiveness and clinical implications This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

75% compliant at end of follow-up JUPITER Inclusion and Exclusion Criteria, Study Flow ?? Prescreened 89,890 Screened 89,863 Screened Reason for Exclusion (%) LDL-C > 130 mg/dL 53 hsCRP < 2.0 mg/L 37 Withdrew Consent 4 Diabetes 1 Hypothyroid <1 Liver Disease <1 TG > 500 mg/dL <1 Age out of range <1 Current Use of HRT <1 Cancer <1 Poor Compliance/Other 3 Men > 50 years Women > 60 years No CVD, No DM LDL < 130 mg/dL hsCRP > 2 mg/L 4 week Placebo Run-In 17,802 Randomized 17,802 Randomized 75% compliant at end of follow-up More patients (26,656) excluded due to CRP <2.0 mg/dL than enrolled in the trial

Caveats Challenging The Interpretation of JUPITER Trial Generalizability to clinical practice questionable - Number of subjects pre-screened not reported - Only one in 5 screened enrolled - Compliant patients enrolled after placebo run-in phase - Only 75% compliance at median follow-up of 2 yrs This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Caveats Challenging The Interpretation of JUPITER Trial Validity of study hypothesis Healthy status of subjects Generalizability to clinical practice Reliability of benefit Mortality benefit Benefit-risk assessment Cost-effectiveness and clinical implications This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Considerations of Stopping a Randomized Clinical Trial Early Unacceptable safety - “primum non nocere” Futility - Poor prospects of a positive trial New external information - Unequivocally establishes efficacy or safety Apparent large benefit - Ethical imperative to protect trial participants This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Randomized Clinical Trial Stopped Early Balancing Contrasting Goals Scientific validity unbiased estimate other outcomes benefit-risk ratio Ethical imperative protect trial participants rapid dissemination This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP. Overly sanguine estimates of treatment effect result in misleading risk-benefit ratios, misguided practice recommendations, and suboptimal clinical practice Montori et al, JAMA 2005; 294:2203–2209 Bassler et al (STOPIT-2), JAMA 2010; 303: 1180-1187

Treatment benefit in JUPITER is unexpectedly large and rapid Was Treatment Benefit Implausibly High? Reconciling JUPITER Results with CTT Meta-analysis CTT meta-analysis 1 mmol/L lowering in LDL-C reduces the risk of vascular events by 10% after 1 year, 25% after 2–3 years, and about 30% after 4 years JUPITER Nearly 50% risk reduction in vascular events in after a 1·2 mmol/L drop in LDL-C after 1·9 years Treatment benefit in JUPITER is unexpectedly large and rapid

Pooled Primary Prevention Statin Trials (N=8 Trials, 42,684 pts) Bayesian Analysis of JUPITER Endpoint Prior RR (95% CI) JUPITER Posterior CORONA CVD, MI or Stroke 0.94 (0.84-1.03) 0.53 (0.40-0.68) 0.87 (0.80-0.94) All cause mortality 0.95 (0.87-1.04) 0.80 (0.67-0.96) 0.92 (0.85-1.00) CHD or nonfatal MI 0.90 (0.79-1.02) 0.51 (0.36-0.72) 0.84 (0.74-0.94) Pooled Primary Prevention Statin Trials (N=8 Trials, 42,684 pts) 0.96 (0.89-1.05) 0.94 (0.86-1.01) 0.74 (0.63-0.87) 0.69 (0.60-0.80) Moderated estimates are substantially less than observed effects

Caveats Challenging The Interpretation of JUPITER Trial Magnitude of treatment benefit not reliable - Impressive relative but not absolute risk differences - Although real, early stopping may have contributed to overestimation of benefit (“too good to be true”) - Details of stopping rules not reported in the primary report - Trials should ideally be stopped for safety or futility concerns, not for “implausibly large” treatment effects - Moderated estimates are substantially less than observed effects This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Caveats Challenging The Interpretation of JUPITER Trial Validity of study hypothesis Healthy status of subjects Generalizability to clinical practice Reliability of benefit Mortality benefit Benefit-risk assessment Cost-effectiveness and clinical implications This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

JUPITER Mortality Data Endpoint Rosuvastatin Placebo HR 95%CI P NNT All-cause death 198 247 0.80 0.67-0.97 0.02 182 Cancer death 35 58 0.61 0.40-0.92 0.02 387 (FDA report) 40 65 0.62 0.42-0.91 0.01 356 Non-cancer death 163 189 0.86 0.70-1.06 0.16 342 Fatal Stroke 3 6 0.54 0.15-1.97 0.35 2967 Fatal MI 9 6 1.46 0.54-3.96 0.46 Fatal Stroke/MI 12 12 1.00 0.46-2.19 1.00 CV death 35 44 0.84 0.51-1.24 NS 989

JUPITER: Secondary Endpoint All-Cause Mortality Placebo 247 / 8901 0.06 ARD 1.1%, 95%CI 0.3%-1.9% ARD 0.7%, 95%CI -0.4%-1.8% - 20 % 0.05 B A 0.04 Cumulative Incidence 0.03 Rosuvastatin 198 / 8901 0.02 0.01 0.00 1 2 3 4 Follow-up (years) Number at Risk Rosuvastatin 8,901 8,847 8,787 6,999 4,312 2,268 1,602 1,192 683 227 Placebo 8,901 8,852 8,775 6,987 4,319 2,295 1,614 1,196 684 246

Caveats Challenging The Interpretation of JUPITER Trial Mortality benefit: Real or an artifact? - Post hoc exploratory secondary endpoint - Mortality not adjudicated (P>0.05 discounting vital status data) - No adjustment for multiple comparison (? Type I error) - Mortality difference not significant when FRS>20% excluded - CV mortality not significantly different - Benefit driven by reduction in cancer death (? real effect) - Convergence of curves at 4.4 yrs c/w 4.0 yrs (“random high”) - Lack of clarity with mortality data questions the wisdom of stopping the trial early This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP. Given these uncertainties, FDA did not allow a mortality claim

Caveats Challenging The Interpretation of JUPITER Trial Validity of study hypothesis Healthy status of subjects Generalizability to clinical practice Reliability of benefit Mortality benefit Benefit-risk assessment Cost-effectiveness and clinical implications This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Investigator reported Incident Diabetes in JUPITER Investigator Reported vs ADA Criterion (Post hoc) Placebo ARD = 3.3% RR = 1.27 (1.17, 1.38) NNH = 31 Rosuvastatin Diabetes incidence (%) ARD = 0.5% RR = 1.27 (1.05, 1.53) NNH = 194 Investigator reported HbA1c >6.5% or FSG >126 mg/dL Diabetes not adjudicated (1000 investigators across 26 countries) Nearly 7-fold increase in risk using the ADA criterion 80% of diabetes occurred in those with fasting glucose >100mg/dL at entry

Benefit-Risk Balance in JUPITER For Every 1000 Patients Treated with Rosuvastatin 12 fewer MACE - 1 fewer CV death - 4 fewer nonfatal MI - 3 fewer nonfatal stroke - 1 fewer hosp. for UA - 3 fewer revascularization 5 incident diabetes (33 based on ADA criterion) 3 liver-related AE (AST >3xULN) 5 myalgia 2 confusional state Increased cost Benefits likely overestimated due to premature stopping Risks underestimated due to early stopping, suboptimal adherence and highly selective population 36 36

Caveats Challenging The Interpretation of JUPITER Trial Informed benefit-risk assessment not possible - Benefit likely overestimated - Risk likely underestimated due to early stopping, suboptimal compliance and highly selective population - Increased risk of incident diabetes in statin arm worrisome - Long-term safety of “ultra-low LDL” levels not known This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Caveats Challenging The Interpretation of JUPITER Trial Validity of study hypothesis Healthy status of subjects Generalizability to clinical practice Reliability of benefit Mortality benefit Benefit-risk assessment Cost-effectiveness and clinical implications This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Ridker et al, Circ Cardiovasc Qual Outcomes 2009;2;616-623 Primary Prevention of CVD (NNT Estimates) Intervention Population Endpoint 5-Yr NNT Rosuvastatin Low LDL, hsCRP Death, MI, Stroke, Revasc. 20 Death, MI, Stroke 29 Pravastatin Hyperlipidemia CHD death, MI 44 Lovastatin CHD death, MI, UA 49 Any coronary event 63 Diuretics Hypertension 86 b blocker Fatal/nonfatal CHD, Stroke 140 Aspirin Men (women) CVD, MI, Stroke 346 (426) Ridker et al, Circ Cardiovasc Qual Outcomes 2009;2;616-623

Individual Components of the Primary Endpoint JUPITER Individual Components of the Primary Endpoint Endpoint Rosuvastatin Placebo HR 95%CI P NNT Primary Endpoint* 142 252 0.56 0.46-0.69 <0.00001 81 Non-fatal MI 22 62 0.35 0.22-0.58 <0.00001 222 Any MI 31 68 0.46 0.30-0.70 <0.0002 241 Non-fatal Stroke 30 58 0.52 0.33-0.80 0.003 318 Any Stroke 33 64 0.52 0.34-0.79 0.002 287 Revascularization or Unstable Angina 76 143 0.53 0.40-0.70 <0.00001 133 MI, Stroke, CV Death 83 158 0.53 0.40-0.68 <0.00001 119 CV death 35 44 0.80 0.51-1.24 0.315 989

Caveats Challenging The Interpretation of JUPITER Trial Cost-effectiveness and implications for practice ICER = $25,198 per QALY (Choudhary et al. JACC, 2011) - Assumptions = 54% RRR in MI - Cost = $3.63/day until 2018, $1/day thereafter - Benefit persisting over 5 years - Low adverse event rate Hlatky (editorial) - “The cost-effectiveness ratio from JUPITER is highly leveraged on the assumption of sustained, deep risk reductions, an assumption for which we have little data." - Cost = $50-95 billion in US (“Is this the best the best use of these healthcare dollars?”) This is what Ridker says: In 2001, in an hypothesis generating analysis of apparently healthy individuals in the AFCAPS / TexCAPS trial*, we observed that those with low levels of both LDL and hsCRP had extremely low vascular event rates and that statin therapy did not reduce events in this subgroup (N=1,448, HR 1.1, 95% CI 0.56-2.08). Thus, a trial of statin therapy in patients with low cholesterol and low hsCRP would not only be infeasible in terms of power and sample size, but would be highly unlikely to show clinical benefit. IS HE CORRECT? In contrast, we also observed within AFCAPS/TexCAPS that among those with low LDL but high hsCRP, vascular event rates were just as high as rates among those with overt hyperlipidemia, and that statin therapy significantly reduced events in this subgroup (N=1,428, HR 0.6, 95% CI 0.34-0.98). The lowest placebo event rates were observed in patients with <median LDL and <median CRP levels (2.2%) and statin treatment in this group had no treatment effect. In contrast, statin produced a significant treatment effect in those with >median CRP and <median LDL as well as >median LDL but <median CRP. However, there is no interaction between treatment effect and CRP groups in patients with > or <median LDL (last column interaction terms). A significant interaction between treatment effect and LDL groups was seen only in patients with <median CRP, but not >median CRP.

Jupiter, not wanting man’s life to be wholly gloomy and grim, has bestowed far more passion than reason you could reckon the ration as twenty-four to one. Moreover, he confined reason to a cramped corner of the head and left all the rest of the body to the passions Erasmus (14th century) Passion to reason in the ratio of 24 to 1!  That equation certainly applies to the “hype” surrounding the JUPITER trial