JOURNAL OF THE AMERICAN COLLEGE OF CARDIOLOGY VOL. 73, NO. 4, 2019 ª 2019 THE AUTHORS. PUBLISHED BY ELSEVIER ON BEHALF OF THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION. THIS IS AN OPEN ACCESS ARTICLE UNDER THE CC BY-NC-ND LICENSE (http://creativecommons.org/licenses/by-nc-nd/4.0/). ORIGINAL INVESTIGATIONS Alirocumab Reduces Total Nonfatal Cardiovascular and Fatal Events The ODYSSEY OUTCOMES Trial Michael Szarek, PHD,a,* Harvey D. White, DSC,b,* Gregory G. Schwartz, MD, PHD,c,* Marco Alings, MD, PHD,d Deepak L. Bhatt, MD, MPH,e Vera A. Bittner, MD, MSPH,f Chern-En Chiang, MD, PHD,g Rafael Diaz, MD,h Jay M. Edelberg, MD, PHD,i Shaun G. Goodman, MD, MSC,j Corinne Hanotin, MD,k Robert A. Harrington, MD,l J. Wouter Jukema, MD, PHD,m Takeshi Kimura, MD,n Robert Gabor Kiss, MD, PHD,o Guillaume Lecorps, MSC,k Kenneth W. Mahaffey, MD,l Angèle Moryusef, MD,i Robert Pordy, MD,p Matthew T. Roe, MD, MHS,q,r Pierluigi Tricoci, MD, PHD,r Denis Xavier, MD, MSC,s Andreas M. Zeiher, MD,t Ph. Gabriel Steg, MD,u,v,* for the ODYSSEY OUTCOMES Committees and Investigatorsy ABSTRACT BACKGROUND The ODYSSEY OUTCOMES (Evaluation of Cardiovascular Outcomes After an Acute Coronary Syndrome During Treatment With Alirocumab) trial compared alirocumab with placebo, added to high-intensity or maximum-tolerated statin treatment, after acute coronary syndrome (ACS) in 18,924 patients. Alirocumab reduced the first occurrence of the primary composite endpoint and was associated with fewer all-cause deaths. OBJECTIVES This pre-specified analysis determined the extent to which alirocumab reduced total (first and subsequent) nonfatal cardiovascular events and all-cause deaths in ODYSSEY OUTCOMES. METHODS Hazard functions for total nonfatal cardiovascular events (myocardial infarction, stroke, ischemia-driven coronary revascularization, and hospitalization for unstable angina or heart failure) and death were jointly estimated, linked by a shared frailty accounting for patient risk heterogeneity and correlated within-patient nonfatal events. An association parameter also quantified the strength of the linkage between risk of nonfatal events and death. The model provides accurate relative estimates of nonfatal event risk if nonfatal events are associated with increased risk for death. RESULTS With 3,064 first and 5,425 total events, 190 fewer first and 385 fewer total nonfatal cardiovascular events or deaths were observed with alirocumab compared with placebo. Alirocumab reduced total nonfatal cardiovascular events (hazard ratio: 0.87; 95% confidence interval: 0.82 to 0.93) and death (hazard ratio: 0.83; 95% confidence interval: 0.71 to 0.97) in the presence of a strong association between nonfatal and fatal event risk. CONCLUSIONS In patients with ACS, the total number of nonfatal cardiovascular events and deaths prevented with alirocumab was twice the number of first events prevented. Consequently, total event reduction is a more comprehensive metric to capture the totality of alirocumab clinical efficacy after ACS. (J Am Coll Cardiol 2019;73:387–96) © 2019 The Authors. Published by Elsevier on behalf of the American College of Cardiology Foundation. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Listen to this manuscript’s audio summary by Editor-in-Chief From the aState University of New York, Downstate School of Public Health, Brooklyn, New York; bUniversity of Auckland and Dr. Valentin Fuster on Green Lane Cardiovascular Services Auckland City Hospital, Auckland, New Zealand; cDivision of Cardiology, University of Col- JACC.org. orado School of Medicine, Aurora, Colorado; dAmphia Ziekenhuis Molengracht, Breda, the Netherlands; eBrigham and Women’s Hospital Heart & Vascular Center and Harvard Medical School, Boston, Massachusetts; fDivision of Cardiovascular Disease, Uni- versity of Alabama at Birmingham, Birmingham, Alabama; gGeneral Clinical Research Center, Taipei Veterans General Hospital and National Yang-Ming University, Taipei, Taiwan; hEstudios Cardiológicos Latinoamérica, Instituto Cardiovascular de Rosario, Rosario, Argentina; iSanofi, Bridgewater, New Jersey; jCanadian VIGOUR Centre, University of Alberta, Edmonton, Alberta, and St. Michael’s Hospital, University of Toronto, Toronto, Ontario, Canada; kSanofi, Paris, France; lStanford Center for Clinical Research, ISSN 0735-1097 https://doi.org/10.1016/j.jacc.2018.10.039 388 Szarek et al. JACC VOL. 73, NO. 4, 2019 Total Event Reduction in ODYSSEY OUTCOMES FEBRUARY 5, 2019:387–96 I ABBREVIATIONS n cardiovascular outcomes trials, the cardiovascular events and all-cause or cause-specific AND ACRONYMS primary efficacy assessment is usually death in patients with stable coronary heart disease based on an intervention delaying the or an acute coronary syndrome (ACS) (1–5); similar ACS = acute coronary syndrome time to first occurrence of an event included findings have been reported with other drug classes in a composite of related nonfatal and fatal (6,7). In trials involving these patient populations, the CI = confidence interval (i.e., death) events. In this setting, patients majority of patients are censored due to surviving the HR = hazard ratio are typically encouraged to remain on ran- follow-up period. An important additional source of LDL-C = low-density lipoprotein cholesterol domized therapy after a first reported censoring that may not be fully appreciated when nonfatal event, such that treatment may evaluating the effect of an intervention on nonfatal continue to modify the risk of subsequent nonfatal events is the occurrence of death, which, unlike other and fatal events. Consequently, an analysis involving types of censoring, prevents both the observation and only the first event may not capture the totality of the occurrence of subsequent nonfatal events. clinical impact of an intervention. Furthermore, the SEE PAGE 397 burden of a disease process may be best assessed by all of the events experienced by a patient, as those If the risks of nonfatal events and death are unre- occurring after the first add to morbidity, mortality, lated to one another, censoring follow-up for nonfatal and health care expenditures. events due to death would be considered “non- Several reports have demonstrated the benefits of informative,” similar to censoring due to completing intensive statin therapy on reducing first and subse- the follow-up period. However, if the risk of nonfatal quent events in composites consisting of nonfatal events is positively associated with the risk of death, Department of Medicine, Stanford University, Stanford, California; mDepartment of Cardiology, Leiden University Medical Center, Leiden, the Netherlands; nKyoto University Graduate School of Medicine, Kyoto-shi, Kyoto, Japan; oMagyar Honvédség Egész- ségügyi Központ, Budapest, Hungary; pRegeneron Pharmaceuticals Inc., Tarrytown, New York; qDivision of Cardiology, Depart- ment of Medicine, Duke University School of Medicine, Durham, North Carolina; rDuke Clinical Research Institute, Duke University Medical Center, Durham, North Carolina; sDepartment of Pharmacology and Division of Clinical Research, St. John’s Medical College and Research Institute, Bangalore, India; tDepartment of Medicine III, Goethe University, Frankfurt am Main, Germany; uAssistance Publique-Hôpitaux de Paris, Hôpital Bichat, Paris and Paris Diderot University, Sorbonne Paris Cité, FACT (French Alliance for Cardiovascular Trials), INSERM U1148, Paris, France; and the vNational Heart and Lung Institute, Imperial College, Royal Brompton Hospital, London, United Kingdom. *Drs. Szarek, White, Schwartz, and Steg contributed equally to this work. yA complete list of the ODYSSEY OUTCOMES Committee members, investigators, and contributors, and their institutional affiliations, is provided in the Online Appendix. This study was supported by Sanofi, Regeneron Pharmaceuticals, Inc., and Fondation Assistance PubliqueHôpitaux de Paris. Dr. Szarek has served as a consultant and/or on the advisory board for CiVi, Resverlogix, Baxter, Esperion, and Regeneron Pharmaceuticals, Inc. Dr. White has received research grants from Sanofi, Eli Lilly, National Institute of Health, George Institute, Omthera Pharmaceuticals, Pfizer New Zealand, Intarcia Therapeutics Inc., Elsai Inc., Dalcor Pharma UK Inc., CSL Behring LLC, and Luitpold Pharmaceuticals Inc.; has received honoraria and nonfinancial support from AstraZeneca; and has served on the advisory boards of Sirtex and Acetilion. Dr. Schwartz has received research support to his institution from Resverlogix, Sanofi, and Roche; and is a co-inventor of pending U.S. patent application 14/657192 (“Methods for Reducing Cardiovascular Risk”) that has been assigned to the University of Colorado. Dr. Alings has received research support from Sanofi and Regeneron Pharmaceuticals; has received honoraria from Pfizer; and has served as a consultant and/or on the advisory board for Bayer, Bristol-Myers Squibb, Boehringer Ingelheim, Milestone, Pfizer, and Daiichi-Sankyo. Dr. Bhatt has served on the advisory board of Cardax, Elsevier Practice Update Cardiology, Medscape Cardiology, and Regado Biosciences; has served on the Board of Directors of Boston VA Research Institute, Society of Cardiovascular Patient Care, and TobeSoft; has served as Chair of the American Heart Association Quality Oversight Committee, NCDR-ACTION Registry Steering Committee, and VA CART Research and Publications Committee; has served on Data Monitoring Committees for Baim Institute for Clinical Research (formerly Harvard Clinical Research Institute, for the PORTICO trial, funded by St. Jude Medical, now Abbott), Cleveland Clinic, Duke Clinical Research Institute, Mayo Clinic, Mount Sinai School of Medicine (for the ENVISAGE trial, funded by Daiichi-Sankyo), and Population Health Research Institute; has received honoraria from American College of Cardiology (senior associate editor, Clinical Trials and News, ACC.org; vice-chair, ACC Accreditation Committee), Baim Institute for Clinical Research (formerly Harvard Clinical Research Institute; RE-DUAL PCI clinical trial steering committee funded by Boehringer Ingelheim), Belvoir Publications (Editor-in-Chief, Harvard Heart Letter), Duke Clinical Research Institute (clinical trial steering committees), HMP Global (Editor-in-Chief, Journal of Invasive Cardiology), Journal of the American College of Cardiology (guest editor; associate editor), Population Health Research Institute (for the COMPASS operations committee, publications committee, steering committee, and U.S. national co-leader, funded by Bayer), Slack Publications (chief medical editor, Cardiology Today’s Intervention), and Society of Cardiovascular Patient Care (secretary/treasurer), and WebMD (CME steering committees); has served as deputy editor of Clinical Cardiology; has received research funding from Abbott, Amarin, Amgen, AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Chiesi, Eisai, Ethicon, Forest Laboratories, Idorsia, Ironwood, Ischemix, Lilly, Medtronic, PhaseBio, Pfizer, Regeneron, Roche, Sanofi, Synaptic, and The Medicines Company; has received royalties from Elsevier (editor, Cardiovascular Intervention: A Companion to Braunwald’s Heart Disease); has served as site co-investigator for Biotronik, Boston Scientific, St. Jude Medical (now Abbott), and Svelte; is a trustee of the American College of Cardiology; and has performed unfunded research for FlowCo, Merck, Novo Nordisk, PLx Pharma, and Takeda. Dr. Bittner has received research grants from Amgen, DalCor, Esperion, Sanofi, AstraZeneca, and Bayer JACC VOL. 73, NO. 4, 2019 Szarek et al. 389 FEBRUARY 5, 2019:387–96 Total Event Reduction in ODYSSEY OUTCOMES the occurrence of death may violate the non- accounting for competing deaths that prevent follow- informative censoring assumption that is integral to up for nonfatal events, resulting in a more accurate statistical methods typically used to analyze total relative estimate (i.e., hazard ratio [HR]) for nonfatal events. This can lead to erroneous estimates of event risk. Our hypothesis was that alirocumab re- nonfatal event risk, and is especially problematic if duces total events following ACS. there is an imbalance in the number of deaths be- tween treatment groups. METHODS As previously reported, when added to high- intensity or maximum-tolerated statin therapy after Details of the study design (9) and primary efficacy ACS, alirocumab reduced the first occurrence of the and safety results (8) have been published. Qualifying primary composite endpoint and was associated with patients were $40 years of age, provided written fewer deaths relative to placebo in the ODYSSEY informed consent, had been hospitalized with OUTCOMES (Evaluation of Cardiovascular Outcomes an ACS (myocardial infarction or unstable angina) After an Acute Coronary Syndrome During Treatment 1 to 12 months prior to randomization, and had With Alirocumab) trial (8). To address the previously a low-density lipoprotein cholesterol (LDL-C) mentioned issues in the analysis of total events, we $70 mg/dl (1.81 mmol/l), nonhigh-density lipo- utilized a novel approach to jointly model total protein cholesterol $100 mg/dl (2.59 mmol/l), or nonfatal cardiovascular and fatal events in a pre- apolipoprotein B $80 mg/dl, measured after specified analysis of the study, allowing for the pos- $2 weeks of stable treatment with atorvastatin 40 to sibility that patients may experience multiple related 80 mg daily, rosuvastatin 20 to 40 mg daily, or the nonfatal events. The method formally quantifies the maximum tolerated dose of either statin (including association between nonfatal events and death while no statin in case of documented intolerance). Healthcare; has received honoraria from the American College of Cardiology, American Heart Association, and National Lipid As- sociation; and has served as a consultant and on the advisory board for Sanofi. Dr. Chiang has received honoraria from Pfizer, Sanofi, Novartis, Merck Sharp and Dohme, AstraZeneca, Daiichi-Sankyo, Bayer, and Boehringer Ingelheim. Dr. Diaz has received honoraria from Sanofi, AstraZeneca, Bayer, and Dalcor. Dr. Edelberg is an employee of Sanofi. Dr. Goodman has received research grants from Daiichi-Sankyo, Luitpold Pharmaceuticals, Merck, Novartis, Servier, Regeneron Pharmaceuticals Inc., Sanofi, Amgen, AstraZeneca, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, CSL Behring, Eli Lilly, Pfizer, and Tenax Therapeutics; has received honoraria from Bristol-Myers Squibb, Eli Lilly, Fenix Group International, Ferring Pharmaceuticals, Merck, Novartis, Pfizer, Servier, Regeneron Pharmaceuticals Inc., Sanofi, Amgen, AstraZeneca, Bayer, and Boehringer Ingelheim; and has served as a consultant and/or on the advisory board for AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Eli Lilly, Pfizer, Servier, Tenax Therapeutics, Sanofi, Amgen, and Bayer. Dr. Hanotin is an employee of Sanofi. Dr. Harrington has received research grants from Apple, CSL, Sanofi, AstraZeneca, Portola, Janssen, Bristol-Myers Squibb, Novartis, and The Medicines Company; has served as a consultant and/or on the advisory board for Amgen, Bayer, Gilead, MyoKardia, and WebMD; and has served on the Board of Directors (unpaid) for the American Heart Association and Stanford HealthCare. Dr. Jukema has received research grants from the Netherlands Heart Foundation, the Interuniversity Cardiology Institute of the Netherlands, and the European Community Framework KP7 Program; and has received other research support from Amgen, Astellas, AstraZeneca, Daiichi-Sankyo, Lilly, Merck- Schering-Plough, Pfizer, Roche, and Sanofi. Dr. Kimura has received research grants from Pfizer, Sanofi, Merck Sharp and Dohme, and Bayer; and has received honoraria from Kowa, Sanofi, Pfizer, Asteras-Amgen-Biopharma, Merck Sharp and Dohme, Bayer, and AstraZeneca. Dr. Lecorps is an employee of and shareholder in Sanofi. Dr. Mahaffey has received research grants from Afferent, Amgen, Apple, AstraZeneca, Cardiva Medical, Inc., Daiichi, Ferring, Google (Verily), Johnson & Johnson, Luitpold, Medtronic, Merck, Novartis, Sanofi, St. Jude, and Tenax; has ownership interest in BioPrint Fitness; and has served as a consultant and/or on the advisory board for Ablynx, AstraZeneca, Baim Institute, Boehringer Ingelheim, Bristol-Myers Squibb, Cardiometabolic Health Congress, Elsevier, GlaxoSmithKline, Johnson & Johnson, Medergy, Medscape, Merck, Mitsubishi, Myokardia, Novartis, Oculeve, Portola, Radiometer, Springer Publishing, Theravance, UCSF, and WebMD. Dr. Moryusef is an employee of Sanofi. Dr. Pordy is an employee of and shareholder in Regeneron Pharmaceuticals, Inc. Dr. Roe has received research grants from American College of Cardiology, American Heart Association, Familial Hypercholesterolemia Foundation, Ferring Pharmaceuticals, Myokardia, Patient Centered Outcomes Research Institute, and Sanofi; has served as a consultant and/or on the advisory board for Amgen, Ardea Biosciences, AstraZeneca, Eli Lilly, and Merck; and has other relationships with Flatiron, Janssen Pharmaceuticals, Novartis, Novo Nordisk, Regeneron Pharmaceuticals, and Roche-Genentech. Dr. Tricoci has received research grants from Merck, Sanofi, and Refeneron; has served as a consultant and/or on the advisory board for Merck; and is an employee of CSL Behring. Dr. Xavier has received research grants and/or meeting support from the National Heart, Lung, and Blood Institute (National Institutes of Health), UK MRC, Wellcome Trust, AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Cadila, Pfizer, Sanofi, Indian Council for Medical Research, Population Health Research Institute, and Duke Clinical Research Institute; and has served on the advisory board for Pfizer. Dr. Zeiher has served as a scientific advisor for Sanofi, Amgen, Pfizer, and Boehringer; and has served as a speaker for Bayer, Novartis, and Vifor. Dr. Steg has received research grants from Amarin, Bayer, Merck, Sanofi, and Servier; and has received speaking or consulting fees from Amarin, Amgen, AstraZeneca, Bayer/Janssen, Boehringer-Ingelheim, Bristol-Myers Squibb, Lilly, Merck, Novartis, Novo Nordisk, Pfizer, Regeneron Pharmaceuticals, Inc., Sanofi, and Servier. Manuscript received September 25, 2018; revised manuscript received October 25, 2018, accepted October 25, 2018. 390 Szarek et al. JACC VOL. 73, NO. 4, 2019 Total Event Reduction in ODYSSEY OUTCOMES FEBRUARY 5, 2019:387–96 heterogeneity and the correlation between nonfatal T A B L E 1 Categories of Nonfatal Cardiovascular Events events within a patient and is also included in the Alirocumab Placebo Total fatal event function. In the latter case, the frailty (n ¼ 9,462) (n ¼ 9,462) (N ¼ 18,924) random effect is multiplied exponentially by an as- Myocardial infarction 866 (39.6) 994 (39.6) 1,860 (39.6) sociation parameter that quantifies the strength of Stroke 131 (6.0) 181 (7.2) 312 (6.6) Unstable angina requiring hospitalization 37 (1.7) 64 (2.5) 101 (2.1) the relationship between the nonfatal and fatal event Heart failure requiring hospitalization 283 (12.9) 276 (11.0) 559 (11.9) processes. Specifically, an association parameter Ischemia-driven coronary 869 (39.8) 998 (39.7) 1,867 (39.7) value equal to 0 indicates that death is non- revascularization procedure informative for nonfatal events, whereas a value Total 2,186 (100.0) 2,513 (100.0) 4,699 (100.0) >0 indicates that patients at greater risk of nonfatal Values are n (%). events are also at greater risk for death. Ignoring informative censoring by death has been shown to yield inaccurate estimates of nonfatal event risk over Randomization in a 1:1 ratio to treatment with alir- time, whereas this joint model has been shown to ocumab 75 mg or matching placebo, stratified by provide accurate relative estimates of nonfatal and country, was performed with 18,924 patients fatal event risk if patients at greater risk of nonfatal meeting study entry criteria. All doses of study events are also at increased risk for death (11). The medication were given by subcutaneous injection Online Appendix provides additional details for the every 2 weeks. model. The primary efficacy endpoint of the study was In its current application, the joint model estimates time to first occurrence of coronary heart disease the effect of alirocumab relative to placebo on total death, nonfatal myocardial infarction, fatal and adjudicated nonfatal cardiovascular events and nonfatal ischemic stroke, or unstable angina separately on all-cause death, as well as the associa- requiring hospitalization. Nonfatal cardiovascular tion between nonfatal cardiovascular events and events recorded in the trial included nonfatal primary death. A semiparametric penalized likelihood tech- endpoints, hemorrhagic stroke, heart failure nique (11) was applied for parameter estimation, us- requiring hospitalization, and ischemia-driven coro- ing splines with 10 knots to estimate baseline nary revascularization. Events included in the hazards, and the shared frailty was assumed to have a primary analysis of the present report were all-cause gamma distribution. Treatment effects on nonfatal death and total nonfatal cardiovascular events as and fatal events are summarized by HRs and corre- defined in the previous text. A sensitivity analysis sponding 95% confidence intervals (CIs), with stan- restricted total nonfatal cardiovascular events to dard errors derived from the final Hessian matrix and myocardial infarction, stroke (including hemorrhag- p values for each estimated effect in the model from ic), or unstable angina requiring hospitalization. z-distributions. Point estimates and corresponding Given the previously reported observation that the 95% CIs and p values were also calculated for the absolute benefit of alirocumab on the study primary association parameters. Note that the estimated efficacy endpoint was greater among patients with treatment HR and 95% CI for all-cause death from a higher LDL-C at study entry, a post hoc analysis joint analysis may differ numerically from that examined possible heterogeneity in the treatment derived by other modeling strategies (e.g., Cox effect on total nonfatal cardiovascular events and regression). deaths in subgroups defined by LDL-C at randomiza- For model convergence purposes, for a given tion ($100 mg/dl vs. <100 mg/dl). All nonfatal car- patient, a nonfatal event that occurred on the same diovascular events and deaths included in the day as death was excluded, and a maximum of analyses were adjudicated by an independent com- 1 nonfatal event was allowed to occur on a given day. mittee blinded to treatment assignment. With these conventions, all nonfatal events and In this analysis, we applied a joint semiparametric deaths within a given patient have distinct event model (sometimes referred to as a frailty model) that times from randomization. allows for multiple nonfatal cardiovascular events Nonparametric mean cumulative function curves within a given patient, while simultaneously assess- were created for total nonfatal cardiovascular events. ing and adjusting for possible informative censoring The mean cumulative function represents the of the nonfatal event process by death. The model expected (i.e., mean) cumulative number of events provides separate hazard functions for nonfatal for a patient at a given point in time after randomi- events and fatal events, linked by a shared frailty (10). zation. For comparative purposes, Kaplan-Meier The frailty random effect accounts for patient risk curves were also created for first nonfatal events JACC VOL. 73, NO. 4, 2019 Szarek et al. 391 FEBRUARY 5, 2019:387–96 Total Event Reduction in ODYSSEY OUTCOMES and plotted with the mean cumulative function randomized to alirocumab had numerically fewer curves. Continuous variables are expressed as median nonfatal cardiovascular events of every type, except (quartile 1, quartile 3), and categorical variables are for heart failure requiring hospitalization. expressed as counts and percentages. Comparisons of Table 2 summarizes baseline characteristics by baseline demographics and clinical characteristics of groups defined by event frequency categories. Pa- patients grouped by categories of nonfatal and fatal tients with at least 1 event were older, had higher event frequencies were by Wilcoxon rank sum tests baseline LDL-C, and were more likely to have for continuous variables and chi-square and Fisher comorbidities than patients without an event during exact tests (where possible) for categorical variables. the study, including diabetes, hypertension, and For all analyses, 2-tailed p values <0.05 were myocardial infarction prior to the ACS index event. considered statistically significant, with no adjust- Comparing groups with at least 1 event, patients with ment for multiple testing. multiple events or an only event of death had higher All analyses were conducted according to baseline LDL-C relative to patients with a single intention-to-treat, including all patients and events nonfatal event, and there were several differences in from randomization to the common study end date terms of comorbidities, including history of chronic (November 11, 2017). Unless otherwise indicated, an- obstructive pulmonary disease, coronary artery alyses were pre-specified prior to unblinding of the bypass graft, or peripheral artery disease. study database. Analyses were performed in SAS The Central Illustration shows the Kaplan-Meier version 9.4 (IBM, Armonk, New York) and R version curves and mean cumulative function plots for first 3.5 (R Foundation, Vienna, Austria). and total nonfatal cardiovascular events, respec- tively, according to treatment group. Based on the RESULTS estimated proportions at 4 years, the risk in both groups and the absolute risk reduction with alir- Patients were followed for survival for a median of 2.8 ocumab was approximately double for total events years (quartile 1, quartile 3: 2.3, 3.4 years), consisting versus first events. Accounting for total events of 27,014 patient-years for the alirocumab group and therefore illustrates the high burden of ongoing dis- 26,915 patient-years for the placebo group. Ascer- ease in the study population and the diminution of tainment was complete for 99.1% and 99.8% of po- that burden by alirocumab. Corresponding (post hoc) tential patient-years of follow-up for nonfatal plots by baseline LDL-C subgroups are presented in cardiovascular events and survival, respectively. Online Figures 1 and 2. Exposure to randomized treatment as a percentage of Table 3 summarizes the distributions of deaths and follow-up for survival was 85.2% and 89.8% for the nonfatal cardiovascular events by ordinal event. alirocumab and placebo groups, respectively; this There were 5,425 total deaths or nonfatal cardiovas- excludes per-protocol blinded exposure to placebo in cular events, 77% greater than first events (n ¼ 3,064). the alirocumab group following 2 consecutive LDL-C The number of patients with a first event includes measurements below 15 mg/dl (9). Among 1,230 1,955 that experienced a primary efficacy endpoint of patients in the alirocumab group and 1,392 patients in the study and 1,109 that experienced a nonfatal car- the placebo group with an initial nonfatal cardiovas- diovascular or fatal event that was not a component cular event, 81.9% (excluding blinded placebo) and of the primary efficacy composite. Furthermore, 84.6%, respectively, were receiving randomized while a majority of patients did not experience an treatment at the time of the event; all but 4 patients event during the study, a sizable subset of patients in the alirocumab group and 3 patients in the placebo experienced >1 event (n ¼ 1,261). Among patients at group continued randomized treatment after the risk for a first event in the alirocumab and placebo nonfatal event. Therefore, consistent with the intent groups, death occurred as a first event in 2.2% and of the study, patients continued their randomized 2.5%, respectively. Notably, conditional on having a treatment beyond their first nonfatal cardiovascular first nonfatal cardiovascular event, the risk of subse- event, thus allowing treatment to potentially influ- quent death was greater. After a first nonfatal car- ence the occurrence of subsequent events. diovascular event occurring an overall median of 1.0 Table 1 summarizes the types and counts of adju- year (quartile 1, quartile 3: 0.4, 1.7 years) after dicated nonfatal cardiovascular events after randomization, death occurred as a second event in randomization. Myocardial infarction and coronary 5.7% and 5.0%, respectively, of the patients in the revascularization were the most common types of alirocumab and placebo groups. Similarly, after a events, and the proportions of each event type within second nonfatal cardiovascular event occurring an the treatment groups were similar. Patients overall median of 1.2 years (quartile 1, quartile 3: 0.6, 392 Szarek et al. JACC VOL. 73, NO. 4, 2019 Total Event Reduction in ODYSSEY OUTCOMES FEBRUARY 5, 2019:387–96 T A B L E 2 Baseline Characteristics of Patients by Category of Number of Nonfatal Cardiovascular and Fatal Events p Value (B) Only (C) Only (D) Multiple (A) No Events Event ¼ Death Event ¼ Nonfatal CV Events (A) vs. (n ¼ 15,860) (n ¼ 442) (n ¼ 1,361) (n ¼ 1,261) (B)þ(C)þ(D) (B) vs. (C) (C) vs. (D) (B) vs. (D) Age, yrs 58 (51, 65) 63 (56, 70) 59 (52, 66) 61 (54, 68) <0.0001 <0.0001 0.0004 <0.0001 Age category <0.0001 <0.0001 0.0003 0.01 <65 yrs 74.5 56.3 71.1 64.2 65 to <75 yrs 20.8 31.5 22.5 26.6 $75 yrs 4.7 12.2 6.4 9.2 Female 24.9 25.1 27.9 26.1 0.03 NS NS NS Region <0.0001 <0.0001 0.0001 <0.0001 Western Europe 22.3 11.3 22.6 22.1 Eastern Europe 29.3 36.0 25.2 22.5 North America 13.8 12.9 20.9 27.4 South America 13.9 20.6 12.3 10.6 Asia 12.8 10.4 9.6 6.3 Rest of world 7.9 8.8 9.4 11.1 Index event <0.0001 0.005 0.02 0.01 NSTEMI 47.4 52.6 52.8 57.8 STEMI 35.6 26.5 32.3 27.2 Unstable angina 17.0 20.8 14.9 15.0 Time from index event to 2.7 (1.7, 4.4) 2.5 (1.7, 3.6) 2.6 (1.7, 4.2) 2.4 (1.6, 3.9) <0.0001 NS 0.03 NS randomization, months Lipid-lowering therapy at <0.0001 NS 0.003 0.004 randomization High dose atorvastatin/ 89.3 88.7 86.5 85.9 rosuvastatin Other LLT 10.0 9.9 12.1 11.2 No LLT 0.7 1.4 1.4 2.9 LDL-C, mg/dl 86 (73, 103) 91 (74, 109) 88 (73, 107) 92 (76, 113) <0.0001 NS 0.0007 NS LDL-C $100 mg/dl 28.6 37.3 32.0 39.3 <0.0001 0.04 <0.0001 NS Diabetes status <0.0001 <0.0001 NS <0.0001 Diabetes 26.8 44.8 33.4 43.4 Pre-diabetes 44.6 34.4 42.0 36.0 Normoglycemia 28.7 20.8 24.5 20.6 Smoking status 0.02 NS NS NS Current 24.0 22.9 24.5 25.5 Former 41.0 41.6 42.1 44.2 Never 35.0 35.5 33.4 30.4 Medical history prior to index event Hypertension 62.4 77.2 73.2 80.7 <0.0001 NS <0.0001 NS Myocardial infarction 16.8 28.5 25.8 39.7 <0.0001 NS <0.0001 <0.0001 Stroke 2.6 7.5 5.7 6.8 <0.0001 NS NS NS Malignant disease 2.6 3.2 3.7 4.9 <0.0001 NS NS NS COPD 3.1 11.3 5.5 10.6 <0.0001 <0.0001 <0.0001 NS CABG 4.2 9.3 8.8 17.0 <0.0001 NS <0.0001 <0.0001 PAD 3.1 8.4 6.6 10.5 <0.0001 NS 0.0003 NS GFR, ml/min per 1.73 m2 78.8 (68.3, 90.6) 73.2 (59.9, 87.3) 76.6 (64.9, 88.6) 74.5 (60.0, 87.3) <0.0001 NS <0.0001 NS GFR <60 ml/min per 1.73 m2 11.9 25.3 17.0 24.6 <0.0001 0.0002 <0.0001 NS Values are median (quartile 1, quartile 3) or column %. NS ¼ p > 0.05. CABG ¼ coronary artery bypass graft; COPD ¼ chronic obstructive pulmonary disease; CV ¼ cardiovascular; GFR ¼ glomerular filtration rate; LDL-C ¼ low-density lipoprotein cholesterol; LLT ¼ lipid-lowering therapy; NSTEMI ¼ non–ST-segment elevation myocardial infarction; PAD ¼ peripheral artery disease; STEMI ¼ ST-segment elevation myocardial infarction. 2.0 years) after randomization, death occurred as a an increased subsequent risk for death. The joint third event in 6.2% and 6.6%, respectively, of the model (Table 4) confirms this observation with an patients in the alirocumab and placebo groups. association parameter of 2.04 (95% CI: 1.78 to 2.29), Qualitatively, these data suggest that each successive linking the risks of nonfatal cardiovascular events prior nonfatal cardiovascular event is associated with and death. An even stronger association (parameter JACC VOL. 73, NO. 4, 2019 Szarek et al. 393 FEBRUARY 5, 2019:387–96 Total Event Reduction in ODYSSEY OUTCOMES C ENTR AL I LL U STRA T I ON Mean Cumulative Functions and Kaplan-Meier Curves for Nonfatal Cardiovascular Events 0.50 Nonfatal Cardiovascular (CV) Events per Patient 0.40 0.30 0.20 0.10 0 0 1 2 3 4 Years Since Randomization Number at Risk Placebo 9,462 9,219 8,888 3,898 737 Alirocumab 9,462 9,217 8,919 3,946 746 Placebo: Total Nonfatal CV Alirocumab: Total Nonfatal CV Placebo: First Nonfatal CV Alirocumab: First Nonfatal CV Szarek, M. et al. J Am Coll Cardiol. 2019;73(4):387–96. Mean cumulative function curves depict the expected total number of nonfatal cardiovascular (CV) events for a given patient in the placebo and alirocumab groups at a given time after randomization. At 4 years, the estimates are 0.357 and 0.301, respectively. In contrast, the expected proportions of patients with a first nonfatal CV event in the placebo and alirocumab groups were 0.183 and 0.160, respectively. estimate 3.29; 95% CI: 2.86 to 3.72) was found be- Table 4 shows that when modeled using total tween death and nonfatal events limited to myocar- nonfatal cardiovascular events, alirocumab treatment dial infarction, stroke, or unstable angina requiring reduced total nonfatal events (HR: 0.87; 95% CI: 0.82 hospitalization. to 0.93) as well as death (HR: 0.83; 95% CI: 0.71 to As depicted in Figure 1, there were 385 fewer total 0.97). Similarly, when modeled using total nonfatal nonfatal cardiovascular or death events with alir- myocardial infarction, stroke, and unstable angina, ocumab (2,905 events for placebo, 2,520 events for alirocumab reduced those events (HR: 0.84; 95% CI: alirocumab), including 190 fewer first nonfatal car- 0.77 to 0.91) and death (HR: 0.82; 95% CI: 0.68 to diovascular or death events (1,627 events for placebo, 0.99). Thus, the inclusion or exclusion of ischemia- 1,437 events for alirocumab) and an additional 195 driven coronary revascularization and hospitaliza- fewer events among the 2,622 patients with a first tion for congestive heart failure had minimal impact nonfatal cardiovascular event. Normalizing for dura- on the estimated relative effects of alirocumab. tion of follow-up, 7.2 first events and 14.6 total events The estimated association parameters were were avoided with alirocumab per 1,000 patient-years considerably >1, indicating that death is informative of assigned treatment. Thus, analysis of first events for the nonfatal cardiovascular event rate. Specif- reflects only about one-half of the total event reduc- ically, conditional on treatment assignment, patients tion associated with alirocumab treatment over a at the highest risk of death were also at elevated median of 2.8 years. risk for nonfatal events, so that death removed 394 Szarek et al. JACC VOL. 73, NO. 4, 2019 Total Event Reduction in ODYSSEY OUTCOMES FEBRUARY 5, 2019:387–96 chronic obstructive pulmonary disorder, history of T A B L E 3 Distributions of Death and Adjudicated Nonfatal Cardiovascular Events by Event Number malignant disease, history of coronary artery bypass graft, history of peripheral artery disease, glomerular Alirocumab Placebo filtration rate <60 ml/min/1.73 m2, and baseline Median Event Median Event Time* Time* LDL-C group (<100 or $100 mg/dl) in both hazard First event functions. Each additional factor was significantly Nonfatal cardiovascular 1,230/9,462 (13.0) 0.9 (0.4, 1.7) 1,392/9,462 (14.7) 1.0 (0.4, 1.8) related (p < 0.05) to risk of nonfatal and/or fatal Death 207/9,462 (2.2) 1.5 (0.7, 2.4) 235/9,462 (2.5) 1.5 (0.8, 2.3) events, and the resulting estimated association Second event parameter of 1.70 (95% CI: 1.44 to 1.96) indicates the Nonfatal cardiovascular 513/1,230 (41.7) 1.2 (0.6, 2.0) 608/1,392 (43.7) 1.3 (0.6, 2.0) linkage between risk of nonfatal and fatal events Death 70/1,230 (5.7) 1.4 (0.7, 2.4) 70/1,392 (5.0) 1.6 (1.0, 2.4) persists even when taking these additional factors Third event into account. Note that geographic region, smoking Nonfatal cardiovascular 188/513 (36.7) 1.6 (1.0, 2.3) 245/608 (40.3) 1.5 (0.9, 2.4) status, and history of hypertension could not be Death 32/513 (6.2) 1.4 (1.0, 2.7) 40/608 (6.6) 1.4 (0.7, 2.4) Fourth and additional entered into the adjusted post hoc model due to event(s) convergence issues. However, in separate post hoc Nonfatal cardiovascular 255 268 models with treatment assignment, the estimated Death 25 47 association parameters for the models with these Total additional characteristics were 2.35 (95% CI: 2.03 Nonfatal cardiovascular 2,186 2,513 Death 334 392 to 2.66), 2.03 (95% CI: 1.78 to 2.29), and 1.97 (95% CI: 1.72 to 2.22), respectively. Values are n/N (%), median (quartile 1, quartile 3), or n. *Median event time is expressed as years since Online Figure 3 displays the total nonfatal cardio- randomization. vascular and death joint model results for the overall study population and for LDL-C subgroups stratified at a baseline level of 100 mg/dl. Among 5,629 patients those patients at highest risk for nonfatal events with baseline LDL-C $100 mg/dl, there were 255 from the risk set. To determine if this association fewer total nonfatal cardiovascular and fatal events would be altered by including additional baseline with alirocumab compared with placebo. Among characteristics of patients expected to be prognostic 13,295 patients with baseline LDL-C <100 mg/dl, there for survival, a post hoc joint model was fit with total were 130 fewer such events with alirocumab than nonfatal cardiovascular events and death with inclu- with placebo. Put another way, 66% of the absolute sion of treatment assignment, age category (<65, event reduction with alirocumab was observed in 65 to <75, or $75 years), diabetes status (diabetes, 30% of the study population defined by baseline prediabetes, or normoglycemia), history of myocar- LDL-C $100 mg/dl. dial infarction prior to the index ACS event, history of DISCUSSION T A B L E 4 Joint Semiparametric Models The ODYSSEY OUTCOMES trial demonstrated that HR (95% CI) p Value adding the PCSK9 monoclonal antibody alirocumab to Death and total nonfatal cardiovascular intensive statin therapy decreases the first occur- events (n ¼ 5,425) rence of major adverse cardiovascular events Alirocumab: placebo HR for nonfatal 0.87 (0.82–0.93) <0.0001 cardiovascular events (n ¼ 2,186 vs. n ¼ 2,513) compared with placebo (8). The present analysis il- Alirocumab: placebo HR for fatal events 0.83 (0.71–0.97) 0.02 lustrates that this treatment effect is magnified when (n ¼ 334 vs. n ¼ 392) total nonfatal cardiovascular events and death are Association between nonfatal cardiovascular – <0.0001 and fatal events: 2.04 (95% CI: 1.78–2.29) considered, with approximately twice as many total Death and total nonfatal MI, stroke, or UA events as first events prevented. Therefore, while the effi- (n ¼ 2,999) cacy of alirocumab treatment after ACS was estab- Alirocumab: placebo HR for nonfatal myocardial infarction, 0.84 (0.77–0.91) <0.0001 stroke, or unstable angina lished on analysis of time to first primary endpoint (n ¼ 1,034 vs. n ¼ 1,239) event, the efficiency of the intervention to reduce Alirocumab: placebo HR for fatal events 0.82 (0.68–0.99) 0.04 morbidity and mortality after ACS, and its benefit to (n ¼ 334 vs. n ¼ 392) Association between nonfatal and fatal events: – <0.0001 reduce the total burden of disease and health care 3.29 (95% CI: 2.86–3.72) costs, are best reflected by an analysis of total events. These findings mirror the pattern observed in prior Frailty variances were statistically significant (p < 0.0001) in both models. CI ¼ confidence interval; HR ¼ hazard ratio. trials of statins or ezetimibe in patients with coronary heart disease or ACS (1–5), indicating the value of JACC VOL. 73, NO. 4, 2019 Szarek et al. 395 FEBRUARY 5, 2019:387–96 Total Event Reduction in ODYSSEY OUTCOMES evaluating any long-term lipid-lowering therapy on F I G U R E 1 First, Subsequent, and Total Nonfatal CV Events and Death by the basis of total event modification. Treatment Group There were more deaths in the placebo group than in the alirocumab group. It can be inferred from the 3,000 2,905 distributions of death and nonfatal cardiovascular events by ordinal event number that experiencing a 2,520 nonfatal cardiovascular event was associated with an –385 Total 1,121 Events increased risk of death, since the incidence of death 2,250 as a second or later event was greater than as a first 956 event. Furthermore, the joint models demonstrated a Number of Events strong statistical association between nonfatal car- 157 diovascular events and death, which was not mean- 1,500 127 ingfully attenuated after accounting for multiple –190 First factors that were prognostic for nonfatal and fatal Events events. Thus, a greater number of “frail” patients in 1,392 the placebo group than the alirocumab were taken out 1,230 750 of the risk set for nonfatal events over time due to the occurrence of death. This includes a greater number of patients in the placebo group (n ¼ 235) than in the 235 207 alirocumab group (n ¼ 207) that died prior to any 0 observed nonfatal events a median of 1.5 years after Placebo Alirocumab randomization. Consequently, the relationship be- Subsequent Event = Nonfatal CV Subsequent Event = Death tween nonfatal and fatal events is an important First Event = Nonfatal CV First Event = Death consideration when interpreting the absolute treat- ment effect on the first event in a composite endpoint There were 385 fewer total nonfatal cardiovascular (CV) or death events with alirocumab that excludes certain causes of death, as well as the (2,905 events for placebo, 2,520 events for alirocumab) versus 190 fewer first nonfatal absolute treatment effect on total events. CV or death events (1,627 events for placebo, 1,437 events for alirocumab). In the previously reported primary analysis of the study data (8), the observed 15% hazard reduction in all-cause death with alirocumab, with p ¼ 0.026 by a patients with baseline LDL-C $100 mg/dl at randomi- stratified log-rank test, was considered nominally zation were less likely to be blindly switched to pla- significant due to the pre-specified testing sequence cebo due to low on-treatment LDL-C (2.3%) of secondary endpoints. The joint models demon- than patients with LDL-C <100 mg/dl at randomiza- strated significant relative reductions in both total tion (10.0%). This may, in part, explain the apparent nonfatal cardiovascular events and death by alir- heterogeneity in the relative treatment effects on ocumab. This complementary modeling strategy total nonfatal cardiovascular events and death. In therefore supports the observation that alirocumab addition, the baseline LDL-C subgroup analyses did reduced all-cause death in the trial. not involve adjustment for other factors that may be STUDY LIMITATIONS. A limitation of the present prognostic for nonfatal cardiovascular events or analysis is the possibility that the apparent relation- death. ship between nonfatal cardiovascular events and death could be explained by other baseline patient CONCLUSIONS characteristics that were not included in the pre- specified or post hoc models. In addition, one Over a median of 2.8 years of follow-up in patients might expect the association between nonfatal with ACS, the total number of nonfatal cardiovascular cardiovascular and fatal events would be restricted events and deaths prevented with alirocumab was to cause-specific deaths (i.e., deaths from cardiovas- twice the number of first events prevented. The pre- cular causes, but not noncardiovascular causes). sent analysis also demonstrated a strong association However, the association parameters in separate between the risks of nonfatal and fatal events during models adjusted for baseline prognostic factors the study. This finding together with the relative re- were statistically significant when fatal events ductions in total nonfatal and fatal events support the were restricted to cardiovascular deaths or non- previously reported observation that alirocumab cardiovascular deaths. Regarding the results for treatment reduced the first occurrence of the primary baseline LDL-C subgroups, it should be noted that composite endpoint and was associated with a 396 Szarek et al. JACC VOL. 73, NO. 4, 2019 Total Event Reduction in ODYSSEY OUTCOMES FEBRUARY 5, 2019:387–96 reduced risk of all-cause death. Given these obser- PERSPECTIVES vations, reduction in total nonfatal and fatal events may be viewed as a preferred metric to summarize the COMPETENCY IN MEDICAL KNOWLEDGE: clinical benefit and efficiency of treatment with Compared with placebo, the PCSK9 inhibitor alirocumab. alirocumab, when added to high-intensity statin ACKNOWLEDGMENTS The authors thank the therapy after an ACS, reduced first and subsequent patients, study coordinators, and investigators who nonfatal cardiovascular events and all-cause mortality participated in this trial; and Sophie Rushton-Smith, over a median of 2.8 years of follow-up. PhD (MedLink Healthcare Communications, Lon- don), for providing editorial assistance in the prepa- TRANSLATIONAL OUTLOOK: Further studies ration of the manuscript (limited to editing for style, are needed to quantify the broader socioeconomic referencing, and figure and table editing). implications of interventions that reduce the total burden of fatal and nonfatal cardiovascular and ADDRESS FOR CORRESPONDENCE: Dr. Michael noncardiovascular events in high-risk patient Szarek, 450 Clarkson Avenue, MS 43, Brooklyn, New populations that accumulate frailty over time. York 11203. E-mail: michael.szarek@downstate.edu. Twitter: @gabrielsteg, @DLBHATTMD. REFERENCES 1. Tikkanen MJ, Szarek M, Fayyad R, et al. Total syndrome: the IMPROVE-IT Trial. J Am Coll Cardiol on long-term cardiovascular outcomes following cardiovascular disease burden: comparing inten- 2016;67:353–61. acute coronary syndromes: rationale and design of sive with moderate statin therapy insights from the ODYSSEY outcomes trial. Am Heart J 2014; 5. Schwartz GG, Fayyad R, Szarek M, DeMicco D, the IDEAL (Incremental Decrease in End Points 168:682–9. Olsson AG. Early, intensive statin treatment re- Through Aggressive Lipid Lowering) trial. J Am duces ’hard’ cardiovascular outcomes after acute 10. Liu L, Wolfe RA, Huang X. Shared frailty Coll Cardiol 2009;54:2353–7. coronary syndrome. Eur J Prev Cardiol 2017;24: models for recurrent events and a terminal event. 2. Murphy SA, Cannon CP, Wiviott SD, McCabe CH, 1294–6. Biometrics 2004;60:747–56. Braunwald E. Reduction in recurrent cardiovascu- 6. Roe MT, Armstrong PW, Fox KA, et al. Prasugrel 11. Rondeau V, Mathoulin-Pelissier S, Jacqmin- lar events with intensive lipid-lowering statin versus clopidogrel for acute coronary syndromes Gadda H, Brouste V, Soubeyran P. Joint frailty therapy compared with moderate lipid-lowering without revascularization. N Engl J Med 2012;367: models for recurring events and death using statin therapy after acute coronary syndromes 1297–309. maximum penalized likelihood estimation: appli- from the PROVE IT-TIMI 22 (Pravastatin or Ator- cation on cancer events. Biostatistics 2007;8: vastatin Evaluation and Infection Therapy– 7. White HD, Huang Z, Tricoci P, et al. Reduction in 708–21. Thrombolysis In Myocardial Infarction 22) trial. overall occurrences of ischemic events with vor- J Am Coll Cardiol 2009;54:2358–62. apaxar: results from TRACER. J Am Heart Assoc 2014;3:1–9. 3. LaRosa JC, Deedwania PC, Shepherd J, et al. KEY WORDS acute coronary syndrome, Comparison of 80 versus 10 mg of atorvastatin on 8. Schwartz GG, Steg PG, Szarek M, et al., for the alirocumab, total events occurrence of cardiovascular events after the first ODYSSEY OUTCOMES Committees and In- event (from the Treating to New Targets [TNT] vestigators. Alirocumab and cardiovascular out- trial). Am J Cardiol 2010;105:283–7. comes after acute coronary syndrome. N Engl J A PPE NDI X For an expanded Methods Med 2018;379:2097–107. 4. Murphy SA, Cannon CP, Blazing MA, et al. section, supplemental figures, and a complete Reduction in total cardiovascular events with 9. Schwartz GG, Bessac L, Berdan LG, et al. Effect list of investigators, please see the online ezetimibe/simvastatin post-acute coronary of alirocumab, a monoclonal antibody to PCSK9, version of this paper.