Department of Epidemiology and Biostatistics DESIGNING CLINICAL RESEARCH Session

1
Department of Epidemiology and Biostatistics
DESIGNING CLINICAL RESEARCHSession 1

• About this course
• Chapters 1 2
• Examples

2
Course Objectives

• 1. Acquire research skills
• 2. Produce a protocol for a study
• 3. Help others in the workshop
• 4. Provide feedback on the workshop
• 5. Have a multiplier effect

3
Course Ingredients

• July 29- Lectures (900 - 950)
• Sept 9 Selected issues from DCR 3 text
• Sections (1000 - 1150)
• Protocol components
• More issues from the text
• Sept 16 5-page protocols due
• Oct 7, 14 Protocol review sessions
• In pairs, new faculty

4
Types of Study

• Not the best choice for this course
• Mice, molecules without humans
• Cost-effectiveness, meta-analysis
• Secondary data analysis
• Qualitative research
• Ideal
• A new observational study or clinical trial
  involving humans ( molecules) that you will do
  this year

5
Computer skills

• You need to know how to
• Word process, use Pubmed
• Use a reference program such as Endnote
• You can learn by
• Getting a mentor or friend to show you
• Taking a course in the UCSF Library

6
Certificate

• For satisfactory performance in all 3 TICR Summer
  Workshop courses, including
• Turning in your 5-page protocol on time
• Turning in your ethics project on time
• Turning in your career plan on time

7
Faculty for sections

• Christian Apfel MD, PhD Anaesthesia
• Heidi Bauer MD, MPH Public Health
• Valerie Flaherman MD, MPH Pediatrics
• Ari Green MD, MAS-CR Neurology
• John Inadomi MD, MPH Gastroenterology
• Steve Hulley MD, MPH Cardiovascular Epidemiology
• Michael Kohn MD, MPP Emergency Medicine
• Kathleen Liu MD, PhD, MAS-CR Nephrology/Pulm Crit
  Care
• Chris Madsen MD, MPH Pediatrics
• Mark Pletcher MD, MPH General Internal Medicine
• Travis Porco PhD Mathematical Modeling
• Joel Simon MD, MPH General Internal Medicine

8
Faculty for sections

• Christian Apfel MD, PhD Clinical Research
  Methods
• Heidi Bauer MD, MPH Clinical Research Methods
• Valerie Flaherman MD, MPH Clinical Research
  Methods
• Ari Green MD, MAS-CR Clinical Research Methods
• John Inadomi MD, MPH Clinical Research Methods
• Steve Hulley MD, MPH Clinical Research Methods
• Michael Kohn MD, MPP Clinical Research Methods
• Kathleen Liu MD, PhD, MAS-CR Clinical Research
  Methods
• Chris Madsen MD, MPH Clinical Research Methods
• Mark Pletcher MD, MPH Clinical Research Methods
• Travis Porco PhD Clinical Research Methods
• Joel Simon MD, MPH Clinical Research Methods

9
Course Coordinator

• Olivia DeLeon
• Olivia_at_epi.ucsf.edu
• 514-8231 (tel)
• 514-8150 (fax)
• (Please let her know if your email address has
  changed by sending her an email)

10
Anatomy of research What its made of

• Research question
• Significance
• Design
• Subjects
• Population
• Sample
• Variables
• Predictor
• Outcome

11
Physiology of research How it works

• Using measurements in a sample
• to draw inferences about
• phenomena in a population

12
(No Transcript)
13
Hulleys Research Question (1993)

• Should postmenopausal women receive hormones?
• Subjects postmenopausal women
• Predictor hormones
• Outcome ?

14
Improved Research Question

• Does estrogen treatment prevent
• heart attacks in postmenopausal women?
• Subjects postmenopausal women
• Predictor estrogen treatment vs none
• Outcome heart attacks

15
Is RQ FINER?

• Feasible
• Interesting
• Novel
• Ethical
• Relevant

16
Is RQ FINER?

• Need to specify the design
• of the study

17
Designs

• Observational study
• Cohort
• Cross-sectional
• Case control
• Randomized clinical trial
• Surrogate endpoints
• Endpoints of primary interest

18
Cohort design

• Subjects
• 5000 post-menopausal women living in the Bay Area
• Predictor
• Taking estrogen?
• Outcome
• Subsequent 5-year incidence of MI

19
Cross-sectional design

• Subjects
• 2000 PM women seen at SFGH
• Predictor
• Taking/took estrogen?
• Outcome
• History of MI?

20
Case-control design

• Subjects
• Cases 50 PM women with MI in the SFGH ED
• Controls 50 PM women with trauma in the SFGH ED
• Predictor
• Taking/took post-menopausal estrogen?
• Outcome
• Cases vs controls

21
Observational Studies of Estrogen and CHD

• Author (year) Relative risk
• Lafferty (1985) 0.2
• Sullivan (1990) 0.2
• Hammond (1979) 0.3
• Nachtigall (1979) 0.3
• Stampfer (1991) 0.3
• Bush (1987) 0.4
• Pettiti (1987) 0.5
• Grodstein (1996) 0.6
• Henderson (1991) 0.7
• Psaty (1994) 0.7
• Wolf (1991) 0.7
• Falkeborn (1992) 0.7
• Criqui (1988) 1.0
• Wilson (1985) 1.9
• Combined 0.7

Barrett-Connor, Public Health Reviews, 1997
p lt .05
22
NB, when choosing a research question and design

• Importance of thorough literature review and
  scholarship

23
Randomized blinded trial design Surrogate
outcomes

• Subjects
• 60 Post-menopausal women
• Predictor
• Randomized to estrogen vs placebo
• Outcome 4 weeks later
• LDL-C decreased by 10, plt.01
• HDL-C increased by 10, plt.01

24
Randomized blinded trial designDisease event
outcomes

• Subjects
• Post-menopausal women
• Predictor
• Randomized to estrogen vs placebo
• Outcome
• Subsequent incidence of MI

25
Feasible?

• Clinical trial of estrogen vs placebo to prevent
  MI/CHD death in 10,000 women with prior
  hysterectomy

26
More feasible

• Secondary prevention trial of estrogen
  progestin vs placebo to prevent MI/CHD death in
  2500 women with a uterus and prior CHD
• Participants willing, available in 20 centers
• Wyeth-Ayerst willing to fund, with UCSF
  controlling the science

27
HERS trial(Heart and Estrogen/progestin
Replacement Study)

• Subjects
• 2763 women age lt 80 (mean age 67)
• postmenopausal, with a uterus
• documented coronary disease
• Predictor
• .625 mg Premarin 2.5 mg MPA (EP)
• vs blinded placebo, randomly assigned
• Outcome
• 4-year rate of non-fatal MI and CHD death

28
Interesting?
29
(No Transcript)
30
Novel?

• First randomized blinded trial with disease
  endpoints of whether estrogen treatment prevents
  CHD

31
Ethical?

• Equipoise (uncertain whether benefits or harms
  predominate)
• Benefits of hormone Rx
• Reduce menopausal symptoms
• ? Prevent CHD
• ? Prevent fractures
• ? Prevent Alzheimers Disease
• ? Improve quality of life
• Harms
• ?Venous thrombo-embolism
• ? Breast cancer

32
Relevant?

• Premarin/Prempro 1 in sales
• Decision faced by half the population

33
Please notice the changes in research question

• Observational RQ
• Is estrogen associated with heart attacks in
  postmenopausal women?
• Intended clinical trial RQ
• Does estrogen prevent CHD events in
  postmenopausal women?
• HERS RQ
• Does estrogen progestin prevent new CHD events
  in postmenopausal women with coronary disease?

34
HERS findings
Treatment Group

• All primary
• CHD events 290 293 .99 .99
• Hulley et al JAMA 1998280605-13

E P
Placebo
RH p
35
Why the null CHD result?Three possibilities

• 1. HERS got the wrong answer
• 2. The observational and other studies got the
  wrong answer
• 3. They answered different questions

36
1. HERS got the wrong answer

• Random error?
• Systematic error?

37
Random error?
Treatment Group
E P
Placebo
95 CI
RH

• All primary
• CHD events 290 293 .99 .84-1.17
• Hulley et al JAMA 1998280605-13

38
HERS got the wrong answer Systematic error?

• Randomization
• Blinding
• Co-intervention
• Biased outcome ascertainment
• Adherence to treatment
• Loss to follow-up

39
2. The observational studies got the wrong answer

• Random error?
• Systematic error?
• Confounding?

40
Confounding

• Big problem in observational studies of drugs for
  preventive medicine
• Women who take hormones are inherently healthier
• Statistical adjustment is only a partial solution
• Only a randomized trial can solve the problem

41
3. HERS answered a different question

• Other populations more responsive?
• Primary prevention earlier in menopause
  
• Other interventions better benefit/harm ratio?
• Estrogen only
• Different E, different P
• Lower doses

42
Three possibilities

• Physiology
• 1. HERS got the wrong answer
• 2. The observational studies got the wrong
  answer
• Anatomy
• 3. They answered different questions

43
How to decide?
44
ReplicationThe primary prevention Womens
Health Initiative

• WHI EP Trial
• Subjects16,608 women with a uterus, mean age 63
• Predictor EP vs placebo (as in HERS)
• Outcome MI CHD death (as in HERS)
• WHI Estrogen-only Trial
• Subjects10,739 women with no uterus, mean age 64
• Predictor E vs placebo
• Outcome MI CHD death

45
Disease outcomes in HERS and WHI
RH (95 CI)

• Outcome HERS EP WHI EP WHI E-alone
• MICHD death 1.0 (0.8-1.2) 1.3 (1.0-1.6) 0.9
  (0.8-1.1)
• Stroke 1.2 (0.9-1.7) 1.4 (1.1-1.8) 1.4 (1.1-1.8)
• Pulm Embolism 2.1 (1.3-3.4) 2.1 (1.6-2.8) 1.3
  (0.9-2.1)
• Breast cancer 1.3 (0.8-1.9) 1.3 (1.0-1.6) 0.8
  (0.6-1.0)
• Hip fracture 1.1 (0.5-2.5) 0.7 (0.5-1.0) 0.6
  (0.4-0.9)
• Dementia 2.0 (1.2-3.5) 1.5 (0.8-2.7)
• Hulley, JAMA 20042911769 (editorial)
• Schumaker, JAMA 20042912947

46
Ethical?

• Equipoise (uncertain whether benefits or harms
  predominate)
• Benefits of hormone Rx
• Reduce menopausal symptoms
• ? Prevent CHD
• ? Prevent fractures
• ? Prevent Alzheimers Disease
• ? Improve quality of life
• Harms
• ?Venous thrombo-embolism
• ? Breast cancer

47
Bottom lines

• HERS did get the right answer
• If properly designed and carried out,
• experiments trump observational studies
• Observational studies of drugs are often
  confounded
• Practice guidelines on hormones after menopause
• Do not use for prevention of CHD, dementia
• This applies to any regimen, pending further
  trials
• Can use for treating menopausal symptoms
• Low dose, short duration

48
Annual Number of US Prescriptions for Hormone
Therapy
WHI
HERS
Hersh, JAMA 200429147
Source IMS Health NPA Plus
49
NB

• HERS and WHI were very large studies.
• How about something bite-sized?

50
The research cycle
Develop research question
Infer conclusions
Design study
Implement study
Analyze results
51
Next
52
One sentence describing anatomy of your study

• Research question
• Design
• Subjects
• Variables
• Predictor
• Outcome

53
FINER?

• Feasible
• Interesting
• Novel
• Ethical
• Relevant