Solving clinical trial problems by using novel designs. Anastasia Ivanova and Sonia Davis-Thomas Department of Biostatistics

Transcription

1 Solving clinical trial problems by using novel designs Anastasia Ivanova and Sonia Davis-Thomas Department of Biostatistics

2 Problem 1 Difficulties with patient recruitment Bias that occurs when patients are aware of a new treatment not available to them and comply poorly with the standard treatment 2

3 The ECMO Circuit ECMO (from Robert Truog s presentation on ethics) 3

4 Randomized Consent (Zelen, 1979; Zelen 1990) Eligible Newborn R A N D O M I Z E Do not seek consent Seek consent for ECMO No Yes CMT CMT ECMO Analysis: intent to treat

5 The Harvard neonatal ECMO trial (O'Rourke et al., 1989): study design Stage I: 50/50 randomization until 4 deaths in one arm Stage II: Assign all patients to the more successful therapy, until 4 deaths in that arm or until statistical significance achieved Seek consent only from those randomized to the experimental therapy ECMO

6 The Harvard neonatal ECMO trial: randomized consent Investigators justifications for Randomized Consent Control patients were not really research subjects Parents of control patients were not really being offered a choice, so why subject them to stress? The response to randomized consent The doctors were doing exactly what physicians did before we had a doctrine of informed consent - making decisions for parents. George Annas, Boston University 6

7 Randomized consent Why randomized consent design can be beneficial? It eliminates a coin toss therefore patient-physician relationship is not compromised. Physicians might refuse to enroll patients in a trial where patients are randomized and one treatment is better than the other. If patients are aware of a new treatment not available to them, they comply poorly with the standard treatment. If patients are aware of a new treatment they might go and seek that new treatment. Randomized consent can be used in trials where patient cannot be blinded, because we have to unblind a patient when we seek their consent. that do not require extensive data collection, as patients will become aware of the trial 7

8 Example: Randomized trial of bone density screening Following screening, those in the lowest quartile of BMD were advised to consider HRT The non-screened group were not contacted at baseline as this might have artificially increased their use of hormone replacement therapy Cancer events for both groups of patients were ascertained through medical records and a cancer registry 8

9 Problem 2: Drop-outs in clinical trials Reasons for Discontinuation in a schizophrenia trial Reason Lack of efficacy Examples More paranoid and could not reason properly Side effects Feels very sedated when waking up Patient Decision Called and said he did not want to take meds Does not believe she is ill and has discontinued all medication 9

10 NAS Missing Data Panel National Academy of Science Panel on Handling Missing Data in Clinical Trials (2010) Recommendation: Investigators, sponsors, and regulators should design clinical trials consistent with the goal of maximizing the number of participants who are maintained on the protocol-specified intervention until the outcome data are collected. 10

11 NAS Missing Data Panel: Design Elements to Minimize Dropouts Treatment Period: short treatment period (or measure primary efficacy early-on and continue following for safety) rescue medications (in combination with time to treatment failure) determine long-term efficacy with randomized withdrawal Nat. Res. Council

12 Design Solution: Multi-sequence 2-period Design A family of designs that addresses both issues: Minimize drop-outs Collect data after drop-out Good design choice for: Chronic diseases Treatments without long-term efficacy effects schizophrenia, bipolar, depression, pain, arthritis, diabetes, HIV, high blood pressure Koch, Davis, Anderson

13 3-Sequence 2-Period Design Sequence Period 1 Period 2 Enhanced Retention A:A Active Active A:P Active Placebo P:A Placebo Active Everyone receiving Placebo in Period 1 will receive Active in Period 2. Some participants receiving Active in Period 1 will stay on Active in Period 2. If discontinue treatment, do not discontinue study. Immediately cross-over to next treatment. Koch, Davis, Anderson

14 Role of Period 1: Primary Efficacy Sequences (AA+AP) vs. PA Sequence Period 1 Period 2 A:A Active Active A:P Active Placebo P:A Placebo Active Randomizatio n Ratio A/P Balance in Period 1 A/P Balance in Period 2 1:1:2 50/50 75/25 1:1:1 66/33 66/33 Koch, Davis, Anderson

15 Role of Period 2 Continued Safety and Efficacy Measures Data collection of key outcomes after primary treatment period AA vs. AP: Randomized Withdrawal Is Period 1 benefit preserved, or is there a loss/reversal? Sequence Period 1 Period 2 A:A Active Active A:P Active Placebo P:A Placebo Active Koch, Davis, Anderson

16 Role of Period 2 AA vs. PA: Delayed Start What is the effect of delaying treatment for 1 period? Note: reduced power and weaker generalizability if many on P do not complete Period 1 Sequence Period 1 Period 2 A:A Active Active A:P Active Placebo P:A Placebo Active Koch, Davis, Anderson

17 Role of Periods 1 and 2 Together AP vs. PA Usual crossover analysis of Active vs. Placebo If no carryover, serves as supportive analysis comparing the 2 treatment groups Sequence Period 1 Period 2 A:A Active Active A:P Active Placebo P:A Placebo Active Koch, Davis, Anderson

18 Example: Tesamorelin for excessive abdominal fat in HIV Sequence 26 weeks 26 weeks T:T Tesamorelin Tesamorelin T:P Tesamorelin Placebo P:T Placebo Tesamorelin Period 1 Primary Re-randomization at beginning of Period 2 for eligible completers Double-blind safety extensions with randomized withdrawal Falutz et al

19 Problem 3: High Placebo Response Yellow pills make the most effective antidepressants, like little doses of pharmaceutical sunshine Red pills can give you a more stimulating kick More is better Placebos taken four times a day deliver greater relief than those taken twice daily. More $ is better Waber, RL, Shiv, B, Carmon, Z, Ariely, D (2008). JAMA 299:

20 3-Sequence 2-Period Design Koch et al 1998 Sequence Period 1 Period 2 A:A Active Active A:P Active Placebo P:A Placebo Active SPCD Fava et al 2003 Sequence Period 1 Period 2 A:A Active Active P:P Placebo Placebo P:A Placebo Active With additional enrichment in period 2 with placebo non-responders Combined analysis of periods 1 and 2 20

21 References Fava, M, Mischoulon D, Iosifescu D, Witte J, Pencina M et al. (2012). A double-blind, placebo-controlled study of Aripiprazole adjunctive to antidepressant therapy among depressed outpatients with inadequate response to prior antidepressant therapy (ADAPT-A Study). Psychotherapy and Psychosomatics 81: Falutz J, Mamputp JC, Potvin D, Moyle G, Soulban G, Longhrey H, Marsolais C, Turner R, Grinspoon S. (2010) Effects of Tesamorelin, a Growth Hormone-Releasing Factor Analog, in Human Immunodeficiency Virus-Infected Patients with Excess Abdominal Fat: A Pooled Analysis of Two Multicenter, Double-Blind Placebo-Controlled Phase 3 Trials with Safety Extension Data The Journal of Clinical Endocrinology & Metabolism. 95(9): Koch GG, Davis SM, Anderson RL. (1998) Methodological advances and plans for improving regulatory success for confirmatory studies. Statistics in Medicine; 17: National Research Council. (2010). The Prevention and Treatment of Missing Data in Clinical Trials. Panel on Handling Missing Data in Clinical Trials. Committee on National Statistics, Division of Behavioral and Social Sciences and Education. Washington, DC: The National Academies Press. Torgerson DJ, Thomas RE, Campbell MK, Reid DM (1997) Randomized trial of osteoporosis screening. Use of hormone replacement therapy and quality-of-life results. Arch Intern Med 157: Zelen M. (1979). A new design for randomized clinical trials. New England Journal of Medicine 300: Zelen M. (1990). Randomized consent designs for clinical trials: An update. Stats in Med 9: