Objectives

1
Objectives

• To discuss study design issues
• Clinical study type
• Sample size
• Protocol

2
Study Designs
Descriptive
Analytic
Experimental
case control
correlational
clinical trial
case report/ case series
cohort
community trial
cross-sectional
3
Experimental Designs

• Experiment a set of observations, conducted
  under controlled circumstances, in which the
  scientist manipulates the conditions to ascertain
  what effect such manipulation has on the
  observations.
• Ideally only one factor is examined (however,
  biological variation exists)
• Clinical Trials (individual in a special
  environment are randomized)
• Field Trials (individuals in the community are
  randomized)
• Community Interventions (whole communities are
  randomized)

4
Field Trials

• Differ from clinical trials in that subjects have
  not yet gotten disease
• (1955) Salk vaccine for Polio
• (1975) Vitamin C in preventing the common cold)
• (1982) MRFIT a field trial of several primary
  preventives of MI (N12,866 and cost 115 million)

5
Community Intervention and Cluster Randomized
Trials

• Community intervention is an extension of a field
  trial that involves intervention on a
  community-wide basis
• (eg. Mass media campaigns)
• (eg. Fluoridated water)
• Cluster randomization - groups of participants
  are randomized. The larger the cluster, the less
  that is accomplished by randomizing.

6
Study Protocol

• Rationale and background
• Objectives
• Study Design
• Inclusion/Exclusion
• Definitions (intervention, measurements,
  adherence)
• Study Flow chart
• Sample Size calculation
• Plan of analysis (interim analysis)
• Appendices
• Questionnaires
• Consent forms
• Instructions to interviewers

7
Example of a flow chart for randomization
8
Example of a comparison table to demonstrate that
randomization was successful
9
Objectives should be stated in terms of an
hypothesis

• Null Hypothesis There is no difference
• Medication A will have not effect on disease
  progression
• Two tailed Hypothesis There is some difference
• Medication A will have some effect on disease
  progression
• One tailed Hypothesis The difference is greater
  or less
• Medication A will reduce deaths due to
  disease X
• Medication A will increase deaths due to
  disease X

10
Hypothesis Testing

• Hypothesis testing involves conducting a test of
  statistical significance and quantifying the
  degree to which sampling variability may account
  for the results observed in a particular study
• When designing data collection tools, keep in
  mind your final analysis
• Statistical Tests ?2 T-test
• Measures of Association
• Odds Ratio, Relative Risk

11
Institutional Review Board

• To assure that the rights and privacy of patients
  is guaranteed
• To assure that patients have given informed
  consent
• Keep consent forms and IRB number handy
• Make sure that you have approval before starting

12
Phases of a Clinical Trial

• Phase I - safety (pharmacokenetics - to determine
  maximum tolerated dose)
• Phase II - Evidence of a response
• Phase III - Safety, efficacy
• Phase IV - Safety, Acceptability, Efficacy

13
Basic Study Design

• Control group
• Random allocation to intervention or no
  intervention (or new treatment v.s. standard of
  care)
• Follow-up to an endpoint

14
Study Diagram - Classic Randomized Controlled
O
Eligible Subject Pool
Int
LTF/C
R
P/SC
O
LTF/C
15
Study Design - Cross-over
E
E
Study Eligibles
R
C
C
16
Types of Designs

• Randomized controlled (classic)
• Cross-over (subject is own control)
• Withdrawl (to assess response to discontinuation
  or reduction)
• Factorial (compares two interventions to control
  or single experiment)
• Hybrid (when historical controls are available)
• Equivalency (to determine if new is as good as
  standard)

17
Response Variables (Outcomes)

• Write the question in advance
• primary response variable should be measurable in
  all subjects
• a subject participation generally ends when the
  primary response variable occurs (unless there
  are combinations of response variables to be
  assessed)
• response variables should be capable of unbiases
  assessments.
• Response variables should be ascertained as
  completely as possible

18
Response or Outcome variables

• You may have outcomes other than hard endpoints
• surrogate markers
• quality of life

19
Recruitment of Subjects

• Generalizability can be compromised if better
  follow-up can be assured.
• Keep a log of who could be recruited and who is
  being recruited to see if adaptations need to
  made.
• Explain study and keep good rapport with persons
  who can help you recruit
• Be realistic when planning
• Dont be too aggressive at recruitment because
  you will suffer on follow-up

20
Purpose of Randomization

• Produce comparable study groups
• Remove investigator bias in allocation of
  subjects
• Guarantees that statistical tests will have valid
  significance levels (probability)

21
Types of Randomization

• Fixed allocation
• Simple randomization
• block randomization
• stratified randomization
• baseline adaptive randomization
• response adaptive randomization

22
Methods for improving data quality

• Study protocol
• clearly defined measures on well constructed data
  collection tools
• training of data collection staff
• continuous assessment of follow-up
• data entry in-range and logical checks
• Monitoring of data collection

23
Blindness

• Bias is a error (systematic or non-systematic) or
  difference between the true value and the
  actually obtained due to all causes other than
  sampling variability
• unblinded, single blinded, double blinded, triple
  blinded

24
Survival Analysis

• This analysis used when subjects are entered over
  a period of time and have various lengths of
  follow-up.
• Dichotomous endpoints
• Kaplan Meier or Product Limit
• Cox Proportional Hazard modeling

25
Intent-to-treat Analysis

• For persons who cross-over to the other arm. You
  classify that person into the arm they were
  originally assigned.
• Less biased results than as treated because you
  maintain randomization.
• Only works if there is not a lot of crossing over
  very early in the study

26
Reasons for withdrawl of Subjects

• Ineligibility (misclassification, imprisonment,
  moved)
• Noncompliance (adverse effects of intervention,
  loss of interest, changes in underlying
  conditions, substance usage)

27
Incidence vs. Prevalence

• In infectious diseases of short duration,
  incidence may be close to prevalence
• In chronic diseases, prevalence will be far
  greater than incidence
• Monitor disease burden by prevalence
• Monitor efficacy of programs by incidence

28
Calculate an Incident Rate
Jan July Jan July Jan July
Jan July Jan July Jan time at
1976 1976 1977 1977 1978 1978 1979 1979 1980
1980 1981 risk Sub A ----------------------

2.0 Sub B -----------------------------
----x 3.0 Sub C
-------------------------------------------------
-------- 5.0 Sub D
---------------------------------------
4.0 Sub E
---------------------------x
2.5 Total Years at risk
16.5
initiation of study ID___cases/___person-
years -- Time followed x development of disease
29
Measures of Associaton

• Since clinical trials are prospective and the
  intervention precedes the outcome, a relative
  risk is calculated.
• Covariates and confounders can be either
  controlled for in the design or adjusted for in
  the analysis

30
Sample Size Calculation

• Prevalence in the control group
• Difference you wish to detect
• Alpha you will accept (.05)
• Power you wish to have (.80)
• Number of controls per cases
• Should include an inflation factor to account for
  loss-to-follow-up

31
Sample Size will increase with

• Decreasing significance levels
• Decreasing ? error (increasing power)
• Decreasing clinical significance (treatment
  differences)
• Increasing variability in the observed data

32
Sample Size and C.I.
O.R. 95 C.I. , P value 1.83 (.74 - 4.55), P lt
.17
O.R. 95 C.I., P value 1.83 (1.38 - 2.44) P lt.
001
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
Example of a clinical trial the MYRIAD study

• Perinatal prophylaxis for prevention of vertical
  transmission of HIV to children became available
  in 1994.
• This protocol require administration of ZDV
  during the third trimester of perinatal period,
  during birth process and to the child after
  birth.
• Since it became available, many children have
  received it and the transmission rate has gone
  dramatically down
• Now there is a push to eradicate perinatal
  transmission

39
Percent of Perinatally HIV Exposed or Infected
Children
who Received or whose Mothers Received any ZDV
Born1993-March 1998 in 29 States, United States
100
80
60
Percent Receiving Zidovudine
40
20
0
1996
1993
1994
1995
1997
1998
N1260
N1313
N1271
N1315
N1291
N281
Quarter-Year of Birth
Any ZDVPrenatal, intrapartum, or neonatal
receipt of Zidovudine to reduce perinatal HIV
transmission
Includes 29 areas that have conducted
pediatric HIV Surveillance since 1993 data
reported through March 1999
40
Perinatally Acquired AIDS Cases by
Quarter-Year of Diagnosis 1985-1998, United
States
250
200
150
Number of Cases
100
50
0
1986
1985
1987
1988
1989
1990
1991
1992
1994
1993
1995
1996
1997
1998
Quarter-Year of Diagnosis
Adjusted for reporting delays and redistribution
of NIRs, data reported through March 1999
41
The Intervention

• To eradicate the disease, researchers much target
  women who come in late for prenatal care and
  dont get access to the first phase of the
  perinatal prophylaxis.
• The multi-centered research group decides to
  adapt DOT (direct observation therapy) and use
  TOT (telephone outreach therapy) to improve
  adherence among women who come in late.
• Women who get randomized to the TOT arm will
  receive a cell phone for the 6 week period and
  will be responsible for calling in every day
• Women on both arms need to come in for 12
  follow-up visits.

42
The study centers
New Orleans
Atlanta
New York
Chicago
Miami
43
The intervention
Mom gives the medicine to the baby 4 times a day
44
The research groups (Domains)
Adherence to medication
Biomedical
Access to care
Barriers to prenatal care
Informed consent
45
The outcome

• The outcome of interest will be if the child
  receives the medication for the 6 week period
• Refinement of the outcome is necessary
• Inclusion/exclusion criteria need to be
  established
• Each of the 5 sites can only provide 20 mom/child
  pairs per year (it is a 5 year study). What kind
  of sample size do we need/can we get?
• What kind of confounders should we consider
• Should incentives be given?
• When should a woman be considered LTF?

46
Non-experimental (analytic) study designs

• Conducted because of ethics, cost or convenience
• Two primary types
• Cohort
• Case-control

47
(No Transcript)
48
(No Transcript)
49
(No Transcript)