Safety, Can You Paradigm? A Statistical Lament Janet Turk Wittes Statistics Collaborative.

Slides:



Advertisements
Similar presentations
Unsafety: Making no mockery of honest ad-hockery Janet Wittes Statistics Collaborative ASA/FDA 2005.
Advertisements

Statins in Renal Failure Andrea Fox Sunnybrook Health Science Center May 2010.
Description of Each Study in the Cross Trial Safety Analysis Solomon SD, et al. Circulation 2008 [Epub Mar 31]
1. 2 The primary Objective of IDEAL LDL-C Simvastatin mg/d Atorvastatin 80 mg/d risk CHD In stable CHD patients IDEAL: The Incremental Decrease.
The INSIGHT study - Reliable blood pressure control and additional benefits for hypertensive patients Anthony M Heagerty Department of Medicine Manchester.
Atrial Fibrillation in Patients with Cryptogenic Stroke Gladstone DJ et al. N Engl J Med 2014; 370: Presented by Kris Huston | July 21, 2014.
Prescribing Information is available at the end of this presentation NHS Surrey Lipid Guidelines Dr Adam Jacques Ashford & St.
TNT: Study Design Treating to New Targets 2 5 years 10,001 Patients Clinically evident CHD LDL-C 130  250 mg/dL following up to 8-week washout and 8-week.
VBWG IDEAL: The Incremental Decrease in End Points Through Aggressive Lipid Lowering Study.
Treating to New Targets Study TNT Trial Presented at The American College of Cardiology Scientific Sessions 2005 Presented by Dr. John C. LaRosa.
1 Therapies Not Indicated Andrew P. DeFilippis, Ty J. Gluckman, James Mudd, Catherine Campbell, Gregg Fonarow & Roger S. Blumenthal.
BEAUTI f UL: morBidity-mortality EvAlUaTion of the I f inhibitor ivabradine in patients with coronary disease and left ventricULar dysfunction Purpose.
An Update on NSAID Labeling and Data Review DSaRM Advisory Committee February 10, 2006 Sharon Hertz, M.D. Deputy Director Division of Anesthesia, Analgesia,
Clinical implications. Burden of coronary disease 56 millions deaths worldwide in millions deaths worldwide in % due to CV disease (~ 16.
Pravastatin in Elderly Individuals at Risk of Vascular Disease Presented at Late Breaking Clinical Trials AHA 2002 PROSPER.
BIOE 301 Lecture Seventeen. Guest Speaker Jay Brollier World Camp Malawi.
Slide Source: Lipids Online Slide Library Women’s Health Initiative: Trial of Estrogen plus Progestin 16,608 women randomized 16,608.
VBWG HPS. Lancet. 2003;361: Gæde P et al. N Engl J Med. 2003;348: Recent statin trials: Reduction in primary outcome in patients with diabetes.
Management of Elevated Cholesterol in the Primary Prevention Group of Adult Japanese (MEGA) Trial MEGA Trial Presented at The American Heart Association.
Placebo-Controls in Short-Term Clinical Trials of Hypertension Sana Al-Khatib, MD, MHS Assistant Professor of Medicine Division of Cardiology Duke University.
Avoiding Cardiovascular Events through COMbination Therapy in Patients LIving with Systolic Hypertension The First Outcomes Trial of Initial Therapy With.
Laura Mucci, Pharm.D. Candidate Mercer University 2012 Preceptor: Dr. Rahimi February 2012.
Incremental Decrease in Clinical Endpoints Through Aggressive Lipid Lowering (IDEAL) Trial IDEAL Trial Presented at The American Heart Association Scientific.
André Lamy Population Health Research Institute Hamilton Health Sciences McMaster University Hamilton, CANADA on behalf of the CORONARY Investigators Disclosures.
Farrokh Alemi, Ph.D..  Compare effectiveness  Low dose rofecoxib (Vioxx)  High dose rofecoxib (Vioxx)  Celecoxib  Other medication.
Aspirin Resistance: Significance, Detection and Clinical Management of This Real Phenomenon Webcast May 10 th, 2004 Sponsored by.
Aim To determine the effects of a Coversyl- based blood pressure lowering regimen on the risk of recurrent stroke among patients with a history of stroke.
VIOXX ™ Gastrointestinal Outcome Research (VIGOR) Arthritis Advisory Committee Meeting February 8, 2001 Lourdes Villalba, M.D. DAAODP, CDER, FDA.
1 Statistics in Drug Development Mark Rothmann, Ph. D.* Division of Biometrics I Food and Drug Administration * The views expressed here are those of the.
Cardiovascular Risk and NSAIDs Arthritis Advisory Committee Meeting November 29, 2006 Sharon Hertz, M.D. Deputy Director Division of Analgesia, Anesthesia,
HOPE: Heart Outcomes Prevention Evaluation study Purpose To evaluate whether the long-acting ACE inhibitor ramipril and/or vitamin E reduce the incidence.
Cardiovascular Risk and NSAIDs Arthritis Advisory Committee Meeting April 12, 2007 Sharon Hertz, M.D. Deputy Director Division of Analgesia, Anesthesia,
Antithrombotic Trialists’ Collaboration An updated collaborative overview of randomised trials of antiplatelet therapy among high-risk patients.
NDAs /772 Etoricoxib Robert B. Shibuya, M.D. Medical Officer Division of Anesthesia, Analgesia, and Rheumatology Products.
The ILLUMINATE Study: Enrollment and Outcomes Philip Barter, et al. N Engl J Med 2007;357:
4S: Scandinavian Simvastatin Survival Study
Clinical Outcomes with Newer Antihyperglycemic Agents FDA-Mandated CV Safety Trials 1.
Hypothesis: baseline risk status of the patients and proximity to a recent cardiovascular event influence the response to dual anti-platelet therapy. Patients.
1 Pulminiq™ Cyclosporine Inhalation Solution Pulmonary Drug Advisory Committee Meeting June 6, 2005 Statistical Evaluation Statistical Evaluation Jyoti.
CAPRICORN Adverse CV Events (Frequency ≥ 1.5%) in Either Treatment Group (Uptitration Phase)
What about VIOXX?. Adenomatous Polyp Prevention on Vioxx (APPROVe) Vioxx (rofecoxib) versus Placebo Basic Clinical Trial Objective: Assess whether Vioxx.
Long-Term Tolerability of Ticagrelor for Secondary Prevention: Insights from PEGASUS-TIMI 54 Trial Marc P. Bonaca, MD, MPH on behalf of the PEGASUS-TIMI.
Long-Term Tolerability of Ticagrelor for Secondary Prevention: Insights from PEGASUS-TIMI 54 Trial Marc P. Bonaca, MD, MPH on behalf of the PEGASUS-TIMI.
Drug Regulation in Controversy: Vioxx November 10, 2004 Sandra L. Kweder, M.D. Deputy Director, Office of New Drugs Center for Drug Evaluation and Research.
Presentation Title R3 이지영 / 김 수 중교 수 님. Introduction Lowering LDL cholesterol levels with statins : Reduce the risk of cardiovascular disease Vascular.
The JUPITER Trial Reference Ridker PM. Rosuvastatin to prevent vascular events in men and women with elevated C-reactive protein. N Engl J Med. 2008;359:2195–2207.
Viagra (sildenafil citrate): Extensive Clinical and Post-Marketing Experience Michael Sweeney, MD Senior Medical Director Pfizer Inc.
CHEST 2013; 144(3): R3 김유진 / Prof. 장나은. Introduction 2  Cardiovascular diseases  common, serious comorbid conditions in patients with COPD cardiac.
Angela Aziz Donnelly April 5, 2016
Presented at the American Diabetes Association
Trial profile SAS denotes the Simvastatin in Aortic Stenosis Study
The Importance of Adequately Powered Studies
LEADER trial: Primary Outcome
HOPE: Heart Outcomes Prevention Evaluation study
Pravastatin in Elderly Individuals at Risk of Vascular Disease
Andre Lamy on behalf of the COMPASS Investigators
CANTOS: The Canakinumab Anti-Inflammatory Thrombosis Outcomes Study
First time a CETP inhibitor shows reduction of serious CV events
SPIRE Program: Studies of PCSK9 Inhibition and the Reduction of Vascular Events Unanticipated attenuation of LDL-c lowering response to humanized PCSK9.
EUCLID Trial design: Patients with peripheral arterial disease (PAD) were randomized to ticagrelor 90 mg twice daily (n = 6,930) vs. clopidogrel 75 mg.
Jane Armitage on behalf of the HPS2-THRIVE Collaborative Group
PRECISION Trial design: Patients with arthritis and increased cardiovascular risk were randomized to celecoxib 100 mg twice daily (n = 8,072) vs. ibuprofen.
Dr. PJ Devereaux on behalf of POISE Investigators
Insights from the Anglo-Scandinavian Cardiac Outcomes Trial (ASCOT)
Dr. PJ Devereaux on behalf of POISE Investigators
(p for noninferiority < 0.001)
ARISE Trial Aggressive Reduction of Inflammation Stops Events
SPIRE Program: Studies of PCSK9 Inhibition and the Reduction of Vascular Events Unanticipated attenuation of LDL-c lowering response to humanized PCSK9.
EMPA-REG OUTCOME: Cumulative incidence of the primary outcome
Basic statistics.
Presentation transcript:

Safety, Can You Paradigm? A Statistical Lament Janet Turk Wittes Statistics Collaborative

2 Harms identified late fenflurmine-phentermine (Fen Phen) Rofecoxib (Vioxx) Troglitazone (Rezulin) HRT(Premarin and PremPro) Celecoxib (Celebrex) Telithromycin (Ketek) Rosiglitazone(Avandia) Antidepressants, anti-epileptics….

3 Could we have identified these harms earlier? Troglitazone (Rezulin) -removed from market in 2000  Lots of liver abnormalities  Severe toxicities noted in 1997  Other equally effective drugs didn’t have same problems

4 Could we have identified these harms earlier? Troglitazone (Rezulin) -removed from market in 2000 Rofecoxib (Vioxx) -removed from market in 2004  Every study showed excess heart attack  Attributed to benefit of naproxen

5 Could we have identified these harms earlier? Troglitazone (Rezulin) – removed from market in 2000 Rofecoxib (Vioxx) -removed from market in 2004 HRT(Premarin/PremPro)-major label change 2006  Heart attacks in Puerto Rican girls on oral contraception -1960’s  Men on estrogens had higher event rates – 1970’s

6 Could we have identified these harms earlier? Troglitazone (Rezulin) – removed from market in 2000 Rofecoxib (Vioxx) -removed from market in 2004 HRT(Premarin and PremPro)-label change 2006 Celecoxib (Celebrex) – paper published 2005 Telithromycin (Ketek) – major label change 2007

7 “CELEBRATE :: CELEBREX” December 2004

8 How we statisticians help to save drugs We find safety boring

9 For efficacy we think hard about…  Outcomes  Population to study  Protocol  Analysis of primary outcome  Control of Type I error rate  Other outcomes  Missing data  Sensitivity analyses

10 How we statisticians save drugs Because we find safety boring….  We don’t look at preclinical and early Phase data  We don’t ask about Chemistry Biology What PK/PD studies show Safety part of analysis plan is an afterthought

11 How the statistical  -police protect drugs We test hypotheses Put events in correct body system Give precise definitions No data dredging Too many type 1 errors if we dredge

12 And we divide and… conquer obfuscate

13 e.g. Neuropathy EventTC Neuropathic pain10 Neuropathy10 Neuropathy NOS52 Neuropathy peripheral20 …21 … …

14 e.g. Neuropathy EventTC … Parathesia32 Parathesia NOS40 Parathesia other01 … Peripheral motor neuropathy60 Peripheral sensory neuropathy32

15 True(ish) data from a coxib C T Cardiac disorders4246 Respiratory3329 Vascular disorders 7 9

16 True(ish) data from a coxib C T Cardiac disorders4246 Angina2 2 Angina aggravated0 2 Angina unstable0 3 … Cardiac arrest0 1 Cardiac failure congest2 0 Coronary artery disease4 7 … Myocardial infarction510

17 True(ish) data from a coxib Respiratory 3329 Dyspnea13 Vascular disorders79 Cerebral infarction01 Pulmonary embolism02 TIA20

18 If you combined… No. of people with at least one serious thromboembolic event or evidence of heart failure PlaceboCoxib 16 27

19 Other ways to save drugs Modified Daley’s Rule: Censor early and often

20 e.g., Rofecoxib- short follow-up

21 Through 36 months

22 With denominators (Bresalier et al. NEJM :1092) (And see Adam Boyd’s poster!)

23 Known or suspected adverse events Monitor them Look at events, their (near) synonyms, labs  Are they real?  Are they too frequent?

24 Hierarchical multiplicity Think of biology Order hierarchy by decreasing  Biological plausibility  Objectivity Look for monotone decreasing hazard ratio

25 Which dose of celecoxib do you want?

26 APC Study (Placebo vs high dose) Outcomen CV death 6 +MI 19 +Stroke 26 +CHF 29 +Angina 34 +CV procedure Other CV62

27 Adenoma Prevention with Celecoxib (APC) Study HR CV death 5.1 +MI 3.8 +Stroke 3.4 +CHF 3.2 +Angina2.1 +CV procedure Other CV1.1

28 APC Study CV death 5.1 ( 0.6, 43.2) +MI 3.8 ( 1.3, 11.4) +Stroke 3.4 ( 1.4, 8.3) +CHF 3.2 ( 1.4, 7.4) +Angina2.1 ( 1.0, 4.3) +CV procedure1.7 ( 1.0, 3.1) Other CV1.1 ( 0.7, 1.8)

29 APC Study CV death 5.1 ( 0.6, 43.2) MI 3.8 ( 1.3, 11.4) Stroke 3.4 ( 1.4, 8.3) CHF 3.2 ( 1.4, 7.4) Angina2.1 ( 1.0, 4.3) CV procedure1.7 ( 1.0, 3.1) Other CV1.1 ( 0.7, 1.8)0.7 Solomon (2006). Circulation 114:1028

30 Unknown harms: usual approach Respond by  Agonizing  Checking informed consent document  Asking for more frequent looks  Asking for more thorough analyses Worry about falsely discovered harm

31 Sentinel events Identify Follow in the next patients Invent formal statistical methods

32 Single sentinel event Childhood vaccine 30 day follow-up for serious adverse events 1 death occurred DSMB: did the vaccine cause the death?

33 Women’s Health Initiative Early in the trial, DSMB noted:  Increase in stroke  Increase in pulmonary embolism  Increase in myocardial infarction Possible sentinel events  Myocardial infarction  The big meanies: stroke, PE, MI

34 Proposal 1.Identify sentinel event (or cluster or rate) 2.Monitor for subsequent occurrence(s)  Have reasonable power  Be statistically unbiased (exclude sentinel)  Type 1 error rate may be large (~0.2) Lachenbruch, Wittes: 2007

35 Safety report sample –abnormal lab values Time A B Total Point [N= 150] [N= 148] [N= 298] _______________________________________________________________ SCREENING RANDOM WEEK WEEK WEEK WEEK WEEK WEEK WEEK ___________________________________________________________________

36 But wait! You also get: Time A B Total Point [N= 150] [N= 148] [N= 298] _______________________________________________________________ SCREENING 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) RANDOM 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 2 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 3 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 4 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 5 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 6 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 7 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 8 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) _______________________________________________________________

37 And 150 pages of where’s Waldo Time A B Total Point [N= 150] [N= 148] [N= 298] _______________________________________________________________ SCREENING 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) RANDOM 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 2 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 3 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 4 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 5 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 6 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 7 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) WEEK 8 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) EARLY TERM 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) UNSCHEDULED 0 (0.00 %) 0 (0.00 %) 0 (0.00 %) _______________________________________________________________

38 And if this isn’t enough… Change from baseline where missing is counted as zero (change in HR=64????) Values out of temporal order Lots and lots of decimal places P-values to 3 and 4 significant digits Etc., etc. etc.

39 We need to change our habits Current statistical approach  One variable at a time  Template applied to all studies  No wonder the docs don’t ask us to work with them! Simple change in attitude  Safety parameters aren’t separable  Focus first from biological insights and previous hints  Then scan the other variables  Then refocus

40 Conclusions Worry about multiplicity, but not too much  Listen to Joe Heyse’s talk this afternoon Beware the censor-happy protocol and analysis Don’t be too much the statistician But don’t forget randomness