Intervention Studies
Sue Lindsay, Ph.D., MSW, MPH
Division of Epidemiology and Biostatistics
Institute for Public Health
San Diego State University
What Causes this Disease?
Risk Factors
•
•
•
•
•
•
•
Age
Gender
Bacteria or virus
Exposure to toxins
Diet
Exercise/activity
Genetics
• Stress
• Childhood trauma
• Body mass index
• Smoking/alcohol
Disease
How can we prevent,
detect, and treat this disease?
Interventions
•
•
•
•
•
•
•
•
•
•
•
•
New screening techniques
New medical technologies
Medication
Surgical interventions
Diet and exercise programs
Educational interventions
Stress reduction programs
Smoking cessation programs
Genetic screenings
Health promotion activities
Case management and support
Others?
Disease
Intervention Studies
Research designs that are utilized
to evaluate the impact of a therapy,
intervention, or prevention strategy
Types of Designs
• Observational Designs (4)
• Comparison Group Designs (4)
• Experimental designs (4)
Clinical Trials (4)
Comparison Group
vs.
Control Group
Observational Designs
1.
Before and after studies
2.
Cross-sectional studies with nonuniform programs
3.
Panel or Cohort Studies
4.
Population-based time series
Observational Designs
• Before and After Studies (pre-test, post-test)
– Select persons to receive the intervention
– Measure outcome of interest before intervention
– Measure outcome of interest after intervention
Intervention
Pre-test
Post-test
Observational Designs
• Cross-Sectional Studies for Non-Uniform Programs
– Select persons to receive the intervention
– Carefully measure participation in the program
– Measure outcome of interest after intervention
High users
Intervention
Low users
Measure
outcome
Observational Designs
• Panel or Cohort Studies
– Select persons to receive the intervention
– Measure participation in the program at multiple points in time
– Measure outcome at multiple points in time
Intervention
Measure
Initial
Measure
6 months
Measure
1 year
Observational Designs
• Population-based time series
– Many cross-sectional measures of a population
– Track the exposure of the population to an intervention
– Compare cross-sectional results with exposure
Measure Intervention
Initial
Survey
Measure Intervention
Measure Intervention
1 year
Survey
2 year
Survey
Comparison Group Designs
1.
Historic or Generic Comparison Group
2.
Concurrent Comparison Group
3.
Matched Comparison Group
4.
Comparison Cross-Over Designs
Comparison Group Designs
• Historic or Generic Comparison Group
– Select persons to receive the intervention
– Select historic control group or publicly available norms
– Measure outcome of interest after intervention
Comparison group:
Previous year
County, State, or National rates
Intervention
Measure
outcome
Comparison Group Designs
• Concurrent Comparison group
– Select persons to receive the intervention
– Select another group that does not receive the intervention
– Measure outcome of interest before and after intervention
Intervention
Comparison Group
No Intervention
Initial
Measure
Final
Measure
Comparison Group Designs
• Matched comparison group
– Select persons to receive the intervention
– Select comparison group matched to intervention group
– Measure outcome of interest before and after intervention
Intervention
Matched comparison group
No Intervention
Initial
Measure
Final
Measure
Comparison Group Designs
• Comparison Cross-Over Designs
– Select persons to receive the intervention
– Select comparison group
– After some time, intervention group becomes comparison group,
and comparison group becomes intervention group.
Intervention
Intervention
No Intervention
No Intervention
Initial
Measure
Measure
At Cross-Over
Final
Measure
Experimental Designs
Clinical Trials
1.
Randomized Clinical Trial
2.
Stratified Randomization
3.
Experimental Cross-Over Designs
4.
Factorial Designs
Clinical Trial Study Designs
Defined
Population
Randomize
Treatment
Treatment
#1
#2
Improved
Not
Improved
Improved
Not
improved
Types of Clinical Trials
(David E. Lilienfeld)
1. Therapeutic Trials
relieve symptoms or improve survivorship
2. Intervention Trials
intervention to prevent disease with persons at risk
3. Preventive (Prophylactic) Trials
determine efficacy of a preventive agent
Selection of Study Subjects
• Criteria for study subject selection must
be very precise and written
• Avoid subjective decision-making concerning
who will be in the study and who will not
Randomization
•
•
•
•
Assignment to either study or control group will not be
predictable
Elimination of investigator bias in the process of selecting
patients for either the treatment or control group.
Randomization increases the likelihood that groups will be
comparable on known variables
Randomization increases the likelihood that groups will be
comparable on unknown variables
Masking (Blinding)
•
Subjects do not know whether they
are in the study or control group
(single blinding)
•
Data collectors and analyzers do not
know which patients are in the study
or control group
(double blinding)
Clinical Trials
• Randomized Clinical Trial
– Select persons to participate in study
– Randomize into intervention (“study”) or control group
– Measure outcome of interest after intervention
Intervention
Randomize participants
into intervention or control group
Control
Frequently uses multiple arms of intervention
Measure
outcome
Measure
outcome
Clinical Trials
• Stratified Randomization
– Stratify variables of interest into subgroups
– Randomize each subgroup into intervention and control groups
– Measure outcome of interest after intervention
Intervention
Control
Randomize Into intervention
or control group
Measure
outcome
Measure
outcome
Clinical Trials
• Experimental Cross-Over Designs
Randomize
– Select persons to receive the intervention
– Randomize them into two different intervention groups
– After some time, the two groups change interventions
Intervention
Intervention
No Intervention
No Intervention
Initial
Measure
Measure
At Cross-Over
Final
Measure
Clinical Trials
• Factorial Designs
– Studying the effects of two independent treatments (A & B)
– Randomize into one of four intervention groups
– Measure outcome of interest after interventions
Treatment A and B
Treatment A only
Randomize participants
into one of four groups
Treatment B only
Neither A nor B
Measure
outcome
Measure
outcome
Measure
outcome
Measure
outcome
Challenges to the Validity of
Clinical Trials
•
Selection of subjects is not well-defined or well followed
•
Improper randomization, failure of randomization procedures
•
Data collection for study subjects and controls not standardized
•
No masking or unintentional unmasking
•
Unintentional cross-over
•
Lack of compliance of study or control subjects
•
Differential drop-out rates
•
Errors in analysis:
• Intent to treat analysis: everyone randomized is analyzed
Analysis of Clinical Trials
• Relative Risk
• Efficacy
• NNT:
Number of patients who would
need to be treated to prevent one
adverse outcome
Relative Risk in a Clinical Trial
Disease
Intervention (+)
Control (-)
a
c
Relative Risk =
No Disease
Risk
b
a
(a + b)*
d
c
(c + d)*
a/(a+b)
c/(c+d)
* Denominator is often person-years in the study
An Example
In a study to evaluate the efficacy of a new treatment for the
common cold, 1,000 children 2-18 years old with newly diagnosed
viral colds were recruited from pediatrician offices. 500 were
randomly assigned to receive the new antiviral therapy and 500
were assigned to a control group. After five days, 200 of the study
group had no cold symptoms, while only 100 of the control group
had no cold symptoms. What is the relative risk of curing the
common cold for the study group compared to the control group.
200/500
100/500
=
0.40
0.20
=
2.0
Children with colds who took the antiviral medication were 2.0
times more likely to by asymptomatic after 5 days than those
who did not.
Efficacy
The reduction in the risk of disease or outcome
because of the intervention.
Rate in the
Control Group
__
Rate in the
Study Group
Rate in the
Control Group
X 100
An Example
In a study to evaluate the efficacy of a new vaccine for the
prevention of a viral respiratory disease (VRD), 1,000 health care
workers who volunteered to work with VRD patients were recruited
by participating hospitals, 500 were given the new vaccine, and 500
received a placebo. After one month’s time, researchers noted that
25 of the treated workers had acquired VRD, while 100 of the
control workers contracted VRD. What is the efficacy of the new
vaccine?
100/500 - 25/500
100/500
0.20-0.05
=
0.20
=
75% of VRD infections can be reduced with the vaccine
0.75
NNT
The number of patients who would need to be
treated (NNT) to prevent one adverse outcome.
1
Rate in the
Untreated Group
_
Rate in the
Treated Group
An Example
In a study to evaluate the effectiveness of a supervised diet and
exercise routine to prevent the development of obesity in overweight
children, 3,000 overweight children were enrolled by participating
pediatricians. 1,500 were randomized to the diet and exercise plan
while 1,500 were randomized to normal pediatric care and
recommendations. After one year’s time, 375 of the normal care
children had developed obesity, while 325 of the diet and exercise
children were obese. Calculate the NNT
1
375/1,500
_
1
325/1,500
=
0.25
_ 0.22
=
33
We would need to treat 33 overweight children to prevent one from becoming obese
A mini quiz
What is the study design?
A pharmaceutical company has developed a new
sunscreen and decides to evaluate it using
lifeguards at Mission Beach. Lifeguards are
recruited and randomly assigned to both an
intervention and control group. Those in the
intervention group are given a lotion that looks
identical to the lotion given to control group. Both
groups are asked to use the sunscreen daily and
report to a spot on the beach each Friday
afternoon to be measured for evidence of sun
damage.
What is the study design?
In an effort to reduce obesity and the potential for
Type II diabetes, many schools are considering
removing high sugar content soda machines from
their campuses. Ten elementary schools in a
school district volunteered to be part of a study to
evaluate this approach and removed soda
machines from their campus. These ten schools
were then matched with schools of similar
socioeconomic status and race/ethnicity with soda
machines. Incoming first graders were weighed
and measured at each school, and followed
forward until sixth grade with measurements each
year.
What is the study design?
Between 1990 and 1995 bicycle safety helmet
laws were enacted in all 50 states, requiring
children to wear helmets at all times while riding
bikes. A study was done using San Diego
Trauma Center data to compare the rates of
bicycle head trauma injuries in the years 1985,
1990, 1995, and 2000.
What is the study design?
The hypothesis is that diet and exercise can both
independently improve the risk of developing
certain types of cancer. Women and men are
recruited to participate in a long-term study of diet
and exercise. Men and women separately are
randomized into one of four groups: 1) a group
educated only about the benefits of good eating
habits, 2) a group educated only about the
benefits of exercise, 3) a group educated about
the benefits of both good eating habits and
exercise, and 4) a group provided no specific
intervention.
What is the study design?
1,000 men being treated for mild hypertension
were recruited to participate in a study of a new
medication for the prevention of mild
hypertension. 500 of the men were randomized
to receive the new medication while 500 were
randomized to continue with their current
medications. Each month, blood pressure
measurements were taken. After six months,
those on the new medication were switched back
to their old medication, while those on their old
medication were switched to the new drug.
What is the study design?
A study was designed to evaluate the use of
home visiting by professional nurses and social
workers for the prevention of family violence.
Intact families reporting stress were assigned a
home visitor who made weekly contacts and
referrals for the family. Comprehensive initial
assessments were performed, and the families
were assessed at six month intervals for family
functioning and evidence of violence.