1 / 46

Design of Clinical Trials Epidemiology 2181

Design of Clinical Trials Epidemiology 2181. Rationale and Hypotheses. September 2, 2004. Sheryl F. Kelsey, Ph.D. H:kelseycoursewk2181lecture 1 rationale. 1. CLINICAL TRIAL. A prospective study comparing the effect and value of intervention(s) against a control in human subjects

lysa
Download Presentation

Design of Clinical Trials Epidemiology 2181

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Design of Clinical TrialsEpidemiology 2181 Rationale and Hypotheses September 2, 2004 Sheryl F. Kelsey, Ph.D. H:\kelsey\coursewk\2181\lecture 1 rationale 1

  2. CLINICAL TRIAL A prospective study comparing the effect and value of intervention(s) against a control in human subjects Planned experiment vs observational study Prospective Intervention: Preventative, diagnostic, screening, drugs, devices, surgical procedures, health service delivery Control: Placebo, standard, alternative Humans Ideal: randomized, double masked (blind) 2 H:\kelsey\coursewk\2181\lecture 1 rationale

  3. ROOTS • Issues related to causal inferences. • “Inference”: drawing conclusions to a population from a sample. • Statistical hypothesis testing • Ethical Issues • “Samaritan” versus “Scientific” ethic • “Do No Harm!” 3 H:\kelsey\coursewk\2181\lecture 1 rationale

  4. CAUSALITY • To show that treatment causes outcome, one needs to show that treatment precedes outcome (easy!), treatment is correlated with outcome (easy!), and that there is no alternative explanation for the association between treatment and outcome. 4 H:\kelsey\coursewk\2181\lecture 1 rationale

  5. ESTABLISH CAUSALITY Define the protocols for treatment and control. Define the population, and sample it. Define a response and how to measure it. Assign at random the sample to treatment and control. Control for bias by ensuring objectivity or masked evaluation. A priori rule as to what “proves” that treatment is better than control. 5 H:\kelsey\coursewk\2181\lecture 1 rationale

  6. ISSUES What is the appropriate control or comparison? When is a placebo control ethical? What is the appropriate population? Dropping subjects after randomization Analysis “by intention to treat.” What if you can’t mask and don’t have objective measures? What if you can’t randomize? “You can’t fix by analysis what you muddle by design.” Light et al. 6 H:\kelsey\coursewk\2181\lecture 1 rationale

  7. RANDOMIZED CLINICAL TRIALS ARE THE “GOLD STANDARD” OF EVALUATION. The “rules” are strategies that have evolved over a century, reacting to repeated errors. If you choose to flout the rules, be prepared for the consequences! 7 H:\kelsey\coursewk\2181\lecture 1 rationale

  8. WOMEN’S HEALTH INIATIVE – HORMONE REPLACEMENT THERAPY (HRT) COMPONENT Four trials: Diet, Calcium & Vitamin D, ERT, PERT July 2002 PERT closed • Unfavorable balance of risks and benefits • Significantly increased risk of breast cancer Background Does HRT prevent heart attacks?

  9. BACKGROUND • Women make estrogen; men don’t • Women’s rate of heart attack lower than men’s … until • Menopause … when they stop producing estrogen • Therefore, giving estrogen will reduce rate of heart attack

  10. EXPERIMENT • Coronary Drug Project (CDP), a study in men, had an estrogen arm • Men on the estrogen arm had increased rate of MI (about 30%) • Arm ended early • Conclusion: estrogen is bad for men, good for women

  11. EPIDEMIOLOGY • Women on HRT have roughly a 50% reduction in rate of heart attack • NON-RANDOMIED • How do we know those on HRT are the “same” as those not on HRT?

  12. SURROGATES • If HRT really reduces rate of heart attack, we should see benefit on risk factors. • PEPI: several different formulations of HRT • Showed benefit on risk factors

  13. WHI QUESTION • For post-menopausal women, does additional of hormones (ERT or PERT) lead to overall benefit? • Decrease in heart attack, hip fracture, colorectal cancer • Increase in breast cancer, pulmonary embolism • No effect on death from other cause

  14. INSTITUTE OF MEDICINE’S ANSWER • The answer is in • The study is expensive, unnecessary, and unethical because it puts women on placebo at unnecessary risk.

  15. DATA SAFETY MONITORING BOARD’S RESPONSE • You, IOM, may know the answer • We, the DSMB, don’t • The study needs to be done to answer a very important question

  16. WOMEN’S HEALTH INITIATIVE (WHI) RESULTS After an average of 5.2 years of follow-up, women who received Prempro (0.625 mg/d conjugated equine estrogens plus 2.5 mg/d medroxyprogesterone) compared with women who received placebo had: • The same rate of mortality • Higher rate of CHD • Higher rate of breast cancer • Higher rate of stroke and pulmonary embolism • Lower rate of colorectal cancer • Lower rate of hip fractures.

  17. HYPOTHESIS Study Population - patient selection – eligibility Treatment - intervention Endpoints - outcome evaluation including timing 17 H:\kelsey\coursewk\2181\lecture 1 rationale

  18. THE HYPOTHESIS MUST BE SPECIFIC • “A is better than B” is not adequate. • Rather “In population W is drug A at a daily dose of Y more efficacious in reducing Z over a period of time T than drug B at a dose of X.” • The specific response variables, Z’s, should be stated in advance as well as estimates of the effects of the intervention 18 H:\kelsey\coursewk\2181\lecture 1 rationale

  19. WHAT IS THE QUESTION? Primary Question primary endpoint Secondary Questions secondary endpoint subgroups alternative follow-up Points Ancillary Questions 19 H:\kelsey\coursewk\2181\lecture 1 rationale

  20. EXAMPLES OF TRIALS TESTING VARIOUS INTERVENTIONS • Prevention • For healthy women in their mid 40’s does an exercise and low fat diet regimen compared to a health education program prevent increase in cholesterol over a 5 year period that accompanies menopause? • WHLP • Screen/diagnosis • PLCO • For healthy people 55 years or older does regular cancer screening compared to usual care reduce long term (16 year) mortality? • Drugs • For HIV + patients does specific dosage of drug as compared with placebo taken over one year prevent weight loss? 20 H:\kelsey\coursewk\2181\lecture 1 rationale

  21. EXAMPLES OF TRIALS TESTING VARIOUS INTERVENTIONS (CONTINUED) Devises For patients in cardiac arrest does a minimally invasive devise for internal heart message compared to standard CPR with external defibrillation reduce 24 hour mortality? Surgical Procedures For post cataract endophthalmitis patients does vitrectomy surgery compared to intravitrial medical therapy result in better vision 9 months later? Health Services Delivery For patients with severe alcoholism is inpatient treatment compared to outpatient treatment more effective in 30 day abstention from alcohol? 21 H:\kelsey\coursewk\2181\lecture 1 rationale

  22. TRIALS OF CURRENT PRACTICE • efficacy and safety • window of opportunity • political environment • cost effectiveness • quality of life • Example: IV antibiotics for patients with endophthalmitis, complication after cataract surgery 22 H:\kelsey\coursewk\2181\lecture 1 rationale

  23. WHY ARE CLINICAL TRIALS NEEDED? Variation among people Variation in course of disease Very few “breakthrough” treatments for chronic disease Make inferences to population who require treatment in the future Scientific method best way to evaluate safety and efficacy of new treatments Can make statements about cause and effect Timing of clinical trials 23 H:\kelsey\coursewk\2181\lecture 1 rationale

  24. WHY RANDOMIZED CONTROLLED TRIALS? Uncontrolled trials Compared to what? wonder cures Example: 20 trials (uncontrolled) 5 FU for bowel cancer 8% to 85% larger: 11% to 55% Much more likely to lead to “success” examples: NEJM ‘53-63’ 50% uncontrolled ‘75-’78 30% uncontrolled 24 H:\kelsey\coursewk\2181\lecture 1 rationale

  25. MORTALITY AND NEUROLOGIC OUTCOME, SIX MONTHS Dead Vegetative Severe Disability Mod.Disability Mild/No Disability N=54 N=746 H:\kelsey\coursewk\2181\lecture 1 rationale 25

  26. HISTORICAL CONTROLS Is the comparison fair? Limitations - Bias Patient selection • criteria not clearly defined • types of patients • more restrictive Experimental environment • data quality – missing values • evaluation of response • ancillary care Historical controls usually exaggerate value of new treatment 26 H:\kelsey\coursewk\2181\lecture 1 rationale

  27. ILLUSTRATION Byar (1976) Prostate Cancer No difference overall but above comparison showed difference Statistical adjustments 5 year recruitment Treatment A Treatment B 27 H:\kelsey\coursewk\2181\lecture 1 rationale

  28. CONCURRENT NON-RANDOMIZED CONTROLS Systematic - can influence who gets what birth date even\odd Judgment Lanarkshire milk experiment Data Banks Effectiveness/Research/Outcomes Research Goldmine or Minefield? 28 H:\kelsey\coursewk\2181\lecture 1 rationale

  29. Data from doctors offices, drugstores, insurance companies, hospitals. Search these computer files to see what treatments work, what does not. (New York Times, Tuesday, August 9, 1994.) Outcomes research is complete rubbish, said Dr. Richard Peto, a statistics professor at Oxford University. “It’s a bit like the Emperor’s new clothes,” he said. “People think they have tons of data and so they must be able to analyze it to see what works. They say they’ll use ‘statistics’ to make adjustments for biases and incompleteness. “I’ve spent more than 20 years working as a statistician and I’ve got a silver medal from the Royal Society, and I can assure you that you cannot use statistics to adjust,” Dr. Peto said. 29 H:\kelsey\coursewk\2181\lecture 1 rationale

  30. Dr. Carl Morris, the Chairman of the statistics department at Harvard University, says that with careful statistical analyses, these biases can be accounted for. It is not easy, he said, and he finds that the methods that most researchers use are flawed. “What worries me is that things will get institutionalized before the best statistical methods are set up,” he said. But Dr. Hillman said, “my biggest concern is that all data are not information, and just because you have a few pieces of data does not mean you can make decisions.” 30 H:\kelsey\coursewk\2181\lecture 1 rationale

  31. HISTORICAL EVENTS IN THE DEVELOPMENT OF CLINICAL TRIALS Year Event • 1747 Experiment with untreated control group (Lind) • 1799 Use of sham procedure (Haygarth) • 1863 Use of placebo treatment • 1923 Application of randomization to experimentation (Fisher) • 1931 Random allocation of treatment to groups of patients (Amberson) • 1937 Start of NIH grant support with creation of the National Cancer Institute • 1946 Nuremberg Code for Human Experimentation • 1948 British Medical Research Council Streptomycin for pulmonary tuberculosis • 1962 Publication of book on clinical trials (Hill ) • 1962 Amendments to Food, Drug and Cosmetic Act of 1938 (United States Congress (Kefauver, Harris) • Publications of U.S. Public Health Services regulations leading to creation of Institutional Review Boards for research involving humans • 1967 Structure for separating the treatment monitoring and treatment administration process (Coronary Drug Project Research Group) • 1979 Establishment of Society for Clinical Trials • 2002 Results of Women’s Health Initiative 31 H:\kelsey\coursewk\2181\lecture 1 rationale

  32. WHY SUCH A SHORT HISTORY? Why such reluctance? reverence for authority physician patient paucity of records lack of facilities lack of active remedies randomization not popular Not intuitively appealing to doctor or patient. 32 H:\kelsey\coursewk\2181\lecture 1 rationale

  33. CHANGE & OPPORTUNITIES • International conference on harmonization • Regulatory changes • NIH funding • Industry role expanding • Health care reform • External pressures • Areas for expansion 33 H:\kelsey\coursewk\2181\lecture 1 rationale

  34. CHANGING RESEARCH CLIMATE The number of contract research organizations growing Number of active investigators increased dramatically Academic research centers losing “market share” continue to lose market share: down from 80% in 1992 to 50% in 1997 Industry willing to trade prestige for efficiency and productivity Since 1990, however, the number of new drug applications (NDA) has increased 200% and the number of patients per NDA has also increased 200% 34 H:\kelsey\coursewk\2181\lecture 1 rationale

  35. INTERNATIONAL CONFERENCE ON HARMONIZATION (ICH) • Europe, North American, Japan • Clinical Study Reports • Statistical Principles H:\kelsey\coursewk\2181\lecture 1 rationale 35

  36. FOOD AND DRUG ADMINISTRATION (FDA) REGULATORY CHANGES • Congressional legislation • Consumer pressures • Balance needs for: • -- Efficiency • -- Timeliness • -- Quality 36 H:\kelsey\coursewk\2181\lecture 1 rationale

  37. EXTERNAL FORCES • Political • Professional lobby • HMO stockholders • Small company investors • Consumer activist groups • -- (e.g. AIDS, Women’s Health) H:\kelsey\coursewk\2181\lecture 1 rationale 37

  38. NIH FUNDING • Recent increase, but may plateau • Increase for breast cancer and AIDS • Quite stable for other cancer sites • Stability similar for other diseases H:\kelsey\coursewk\2181\lecture 1 rationale 38

  39. ACADEMIC/INDUSTRY GOVERNMENT ROLES • Many more industry sponsored trials • (Lancet, 1997) • -- cardiology, AIDS, cancer • Industry may focus on drugs and biologics • for therapy • NIH may focus more on • -- prevention, screening, procedures • Medicare pay for clinical trial intervention H:\kelsey\coursewk\2181\lecture 1 rationale 39

  40. GENETICS • New technology allows potential to: • Identify high risk patients • Confirm diagnosis • Identify treatments • Track treatment effects (e.g. surrogates) • Ethical Challenges 40 H:\kelsey\coursewk\2181\lecture 1 rationale

  41. SCREENING AND DIAGNOSTICS • Even more important in future • Not many trials • Ethically challenging • (e.g., Newborn Cystic Fibrosis Screening • Trial) • Results can be controversial • (e.g., Mammography) 41 H:\kelsey\coursewk\2181\lecture 1 rationale

  42. HEALTH MAINTENANCE ORGANIZATIONS • Play major role in health care delivery • Must be cost conscious but consider clinical effectiveness (therapeutic, diagnostic, prevention) • Clinical Trials offers methodology to evaluate cost/effective strategies • Network of HMO’s H:\kelsey\coursewk\2181\lecture 1 rationale 42

  43. BEHAVIORAL MODIFICATION • Most of health care depends on population behavior • Prevention, diagnosis, therapeutics • Still not very effective in population behavior • -- Example: smoking and cancer • Behavioral health research challenges 43 H:\kelsey\coursewk\2181\lecture 1 rationale

  44. COMPLEMENTARY AND ALTERNATIVE MEDICINE • In 1990, 34% of Americans used unconventional therapy at a cost of $13.7 billion, $10.3 billion out of pocket • In 1997, 42% $21.2 billion, $12.2 billion • Much of these therapies untested • Food supplements do not require clinical trial evidence of benefit as long as do not claim prevent or cure disease • Special challenges for clinical trials

  45. TRAINING NEXT GENERATION • Biostatisticians in short supply with • -- Technical skills • -- Clinical trial methodology • -- Communication skills • Research CT physicians also in short supply • -- Little formal CT training • -- No standards to conduct trials • -- Need some level of training and certification 45 H:\kelsey\coursewk\2181\lecture 1 rationale

  46. Alzheimers Disease Clinical Trials (1999) 46

More Related